Download textbook Mathematical And Engineering Methods In Computer Science 9Th International Doctoral Workshop Memics 2014 Telc Czech Republic October 17 19 2014 Revised Selected Papers 1St Edition Petr Hlineny ebook all chapter pdf

Mathematical and Engineering Methods
in Computer Science 9th International

Doctoral Workshop MEMICS 2014 Tel■
Czech Republic October 17 19 2014
Revised Selected Papers 1st Edition
Petr Hlin■ný
Visit to download the full and correct content document:
https://textbookfull.com/product/mathematical-and-engineering-methods-in-computer-
science-9th-international-doctoral-workshop-memics-2014-telc-czech-republic-octobe
r-17-19-2014-revised-selected-papers-1st-edition-petr-hlineny/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Combinatorial Algorithms 25th International Workshop

IWOCA 2014 Duluth MN USA October 15 17 2014 Revised
Selected Papers 1st Edition Kratochvíl Jan
https://textbookfull.com/product/combinatorial-algorithms-25th-
international-workshop-iwoca-2014-duluth-mn-usa-
october-15-17-2014-revised-selected-papers-1st-edition-
kratochvil-jan/
Information Search Integration and Personalization 9th

International Workshop ISIP 2014 Kuala Lumpur Malaysia
October 9 10 2014 Revised Selected Papers 1st Edition
Dimitrios Kotzinos
https://textbookfull.com/product/information-search-integration-
and-personalization-9th-international-workshop-isip-2014-kuala-
lumpur-malaysia-october-9-10-2014-revised-selected-papers-1st-
edition-dimitrios-kotzinos/
Semantic Web Collaborative Spaces Second International

Workshop SWCS 2013 Montpellier France May 27 2013 Third
International Workshop SWCS 2014 Trentino Italy October
19 2014 Revised Selected and Invited Papers 1st Edition
Pascal Molli
https://textbookfull.com/product/semantic-web-collaborative-
spaces-second-international-workshop-swcs-2013-montpellier-
france-may-27-2013-third-international-workshop-
swcs-2014-trentino-italy-october-19-2014-revised-selected-and-
invi/
Digital Forensics and Watermarking 13th International
Workshop IWDW 2014 Taipei Taiwan October 1 4 2014
Revised Selected Papers 1st Edition Yun-Qing Shi
https://textbookfull.com/product/digital-forensics-and-
watermarking-13th-international-workshop-iwdw-2014-taipei-taiwan-
october-1-4-2014-revised-selected-papers-1st-edition-yun-qing-
Engineering Multi Agent Systems Second International
Workshop EMAS 2014 Paris France May 5 6 2014 Revised
Selected Papers 1st Edition Fabiano Dalpiaz
https://textbookfull.com/product/engineering-multi-agent-systems-
second-international-workshop-emas-2014-paris-france-
may-5-6-2014-revised-selected-papers-1st-edition-fabiano-dalpiaz/
Medical Computer Vision Algorithms for Big Data

International Workshop MCV 2014 Held in Conjunction
with MICCAI 2014 Cambridge MA USA September 18 2014
Revised Selected Papers 1st Edition Bjoern Menze
https://textbookfull.com/product/medical-computer-vision-
algorithms-for-big-data-international-workshop-mcv-2014-held-in-
conjunction-with-miccai-2014-cambridge-ma-usa-
september-18-2014-revised-selected-papers-1st-edition-bjoern-
menze/
Progress in Cryptology LATINCRYPT 2014 Third
International Conference on Cryptology and Information
Security in Latin America Florianópolis Brazil
September 17 19 2014 Revised Selected Papers 1st
Edition Diego F. Aranha
https://textbookfull.com/product/progress-in-cryptology-
latincrypt-2014-third-international-conference-on-cryptology-and-
information-security-in-latin-america-florianopolis-brazil-
september-17-19-2014-revised-selected-papers-1st-edit/
Biometric Authentication First International Workshop

BIOMET 2014 Sofia Bulgaria June 23 24 2014 Revised
Selected Papers 1st Edition Virginio Cantoni
https://textbookfull.com/product/biometric-authentication-first-
international-workshop-biomet-2014-sofia-bulgaria-
june-23-24-2014-revised-selected-papers-1st-edition-virginio-
cantoni/
Parameterized and Exact Computation 9th International

Symposium IPEC 2014 Wroclaw Poland September 10 12 2014
Revised Selected Papers 1st Edition Marek Cygan
https://textbookfull.com/product/parameterized-and-exact-
computation-9th-international-symposium-ipec-2014-wroclaw-poland-
september-10-12-2014-revised-selected-papers-1st-edition-marek-
Petr Hlinený · Zdenek Dvorák
Jirí Jaroš · Jan Kofron · Jan Korenek
Petr Matula · Karel Pala (Eds.)
Mathematical and
LNCS 8934
Engineering Methods
in Computer Science
9th International Doctoral Workshop, MEMICS 2014
Telč, Czech Republic, October 17–19, 2014
Revised Selected Papers
123
Lecture Notes in Computer Science 8934
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zürich, Zürich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7408
Petr Hliněný · Zdeněk Dvořák
Jiří Jaroš · Jan Kofroň
Jan Kořenek · Petr Matula
Karel Pala (Eds.)
Mathematical and
Engineering Methods
in Computer Science
9th International Doctoral Workshop, MEMICS 2014
Telč, Czech Republic, October 17–19, 2014
Revised Selected Papers
ABC
Editors
Petr Hliněný Jan Kořenek
Masaryk University Brno University of Technology
Brno Brno
Czech Republic Czech Republic
Zdeněk Dvořák Petr Matula

Charles University Masaryk University
Praha Brno
Jiří Jaroš Karel Pala

Brno University of Technology Masaryk University
Brno Brno
Jan Kofroň
Charles University
Praha
Czech Republic
ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-14895-3 ISBN 978-3-319-14896-0 (eBook)
DOI 10.1007/978-3-319-14896-0
Library of Congress Control Number: 2014960061
LNCS Sublibrary: SL2 – Programming and Software Engineering
Springer Cham Heidelberg New York Dordrecht London

c Springer International Publishing Switzerland 2014
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broad-
casting, reproduction on microfilms or in any other physical way, and transmission or information storage
and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known
or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made.
Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media
(www.springer.com)
Preface
This volume contains the proceedings of the 9th Doctoral Workshop on Mathematical
and Engineering Methods in Computer Science (MEMICS 2014) held in Telč, Czech
Republic, during October 17–19, 2014.
The aim of the MEMICS workshop series is to provide an opportunity for PhD
students to present and discuss their work in an international environment. The scope
of MEMICS is broad and covers many fields of computer science and engineering.
In the year 2014, submissions were invited especially in the following (though not ex-
clusive) areas:
– Algorithms, logic, and games,
– High performance computing,
– Computer-aided analysis, verification, and testing,
– Hardware design and diagnostics,
– Computer graphics and image processing, and
– Artificial intelligence and natural language processing.
There were 28 submissions from PhD students from 10 countries. Each submis-
sion was thoroughly evaluated by at least four Program Committee members who also
provided extensive feedback to the authors. Out of these submissions, nine full papers
were selected for publication in these post-proceedings, and additional nine papers were
selected for local presentation at the workshop.
The highlights of the MEMICS 2014 program included six keynote lectures de-
livered by internationally recognized researchers from the aforementioned areas of
interest. The speakers were:
– Gianni Antichi from University of Cambridge who gave a talk on “An Open-Source
Hardware Approach for High Performance Low-Cost QoS Monitoring of VoIP
Traffic,”
– Derek Groen from University College London who gave a talk on
“High-Performance Multiscale Computing for Modelling Cerebrovascular Blood-
flow and Nanomaterials,”
– Jozef Ivanecký from European Media Laboratory who gave a talk on “Today’s
Challenges for Embedded ASR,”
– Daniel Lokshtanov from University of Bergen who gave a talk on “Tree Decompo-
sitions and Graph Algorithms,”
– Michael Tautschnig from Queen Mary University of London who gave a talk on
“Automating Software Analysis at Large Scale,” and
– Stefan Wörz from University of Heidelberg who gave a talk on “3D Model-Based
Segmentation of 3D Biomedical Images.”
The full papers of four of these keynote lectures are also included in the proceedings.
In addition to regular papers, MEMICS workshops traditionally invite PhD students
to submit a presentation of their recent research results that have already undergone a
VI Preface
rigorous peer-review process and have been presented at a high quality international
conference or published in a recognized journal. A total of 16 presentations out of
22 submissions from 6 countries were included into the MEMICS 2014 program.
The MEMICS tradition of best paper awards continued also in the year 2014. The
best contributed papers were selected during the workshop, taking into account their
scientific and technical contribution together with the quality of presentation. The 2014
awards went to the following papers:
– “LTL Model Checking of LLVM Bitcode with Symbolic Data” by Petr Bauch,
Vojtěch Havel, and Jiří Barnat; presented by Petr Bauch.
– “A New Concept in Advice Complexity of Job Shop Scheduling” by David Wehner;
presented by the author.
The two awards consisted of a diploma accompanied by a financial prize of 400 Euro
each. The prize money was donated by Red Hat Czech Republic and by Y Soft, two
of the MEMICS 2014 Industrial Sponsors.
The successful organization of MEMICS 2014 would not be possible without gen-
erous help and support from the organizing institutions: Brno University of Technology
and Masaryk University in Brno.
We thank the Program Committee members and the external reviewers for their
careful and constructive work. We thank the Organizing Committee members who
helped to create a unique and relaxed atmosphere which distinguishes MEMICS from
other computer science meetings. We also gratefully acknowledge the support of the
EasyChair system and the great cooperation with the Lecture Notes in Computer Sci-
ence team of Springer Verlag.
December 2014 Petr Hliněný

Zdeněk Dvořák
Jiří Jaroš
Jan Kofroň
Jan Kořenek
Petr Matula
Karel Pala
Organization
Workshop
The 9th Doctoral Workshop on Mathematical and Engineering Methods in Computer
Science (MEMICS 2014) was held in Telč, Czech Republic, during October 17–19,
2014. The workshop was hosted by the Telč University Centre of Masaryk University
(former Jesuits College), in a small and pleasant town of Telč inscribed in the UN-
ESCO List. The workshop was attended by 70 participants. More information about the
MEMICS workshop series is available at http://www.memics.cz/.
General Chair
Petr Hliněný Masaryk University, Czech Republic
Program Committee Track Co-chairs

Zdeněk Dvořák Charles University, Czech Republic
Jiří Jaroš Brno University of Technology, Czech Republic
Jan Kofroň Charles University, Czech Republic
Jan Kořenek Brno University of Technology, Czech Republic
Petr Matula Masaryk University, Czech Republic
Karel Pala Masaryk University, Czech Republic
Program Committee
Gianni Antichi University of Cambridge, UK
Tomáš Brázdil Masaryk University, Czech Republic
Markus Chimani Osnabrück University, Germany
Jan Černocký Brno University of Technology, Czech Republic
Eva Dokladalova ESIEE Paris, France
Jiří Filipovič Masaryk University, Czech Republic
Robert Ganian Vienna University of Technology, Austria
Dieter Gollmann TU Hamburg, Germany
Derek Groen University College London, UK
Juraj Hromkovič ETH Zürich, Switzerland
Ondej Jakl VŠB-TU Ostrava, Czech Republic
Hidde de Jong Inria, France
Zdeněk Kotásek Brno University of Technology, Czech Republic
Lukasz Kowalik University of Warsaw, Poland
VIII Organization
Hana Kubátová Czech Technical University in Prague,

Czech Republic
Michal Laclavík Slovak Academy of Sciences, Slovakia
Markéta Lopatková Charles University in Prague, Czech Republic
Julius Parulek University of Bergen, Norway
Maciej Piasecki Wrocław University of Technology, Poland
Geraint Price Royal Holloway, University of London, UK
Viktor Puš CESNET, Czech Republic
Ricardo J. Rodríguez Technical University of Madrid, Spain
Adam Rogalewicz Brno University of Technology, Czech Republic
Cristina Seceleanu MDH, Sweden
Jiří Srba Aalborg University, Denmark
Andreas Steininger TU Wien, Austria
Jan Strejček Masaryk University, Czech Republic
David Šafránek Masaryk University, Czech Republic
Ivan Šimeček Czech Technical University in Prague,
Czech Republic
Petr Švenda Masaryk University, Czech Republic
Catia Trubiani GSSI, Italy
Pavel Zemčík Brno University of Technology, Czech Republic
Florian Zuleger TU Wien, Austria
Steering Committee
Tomáš Vojnar, Chair Brno University of Technology, Czech Republic
Milan Češka Brno University of Technology, Czech Republic
Zdeněk Kotásek Brno University of Technology, Czech Republic
Mojmír Křetínský Masaryk University, Czech Republic
Antonín Kučera Masaryk University, Czech Republic
Luděk Matyska Masaryk University, Czech Republic
Organizing Committee
Radek Kočí, Chair Brno University of Technology, Czech Republic
Zdeněk Letko Brno University of Technology, Czech Republic
Jaroslav Rozman Brno University of Technology, Czech Republic
Hana Pluháčková Brno University of Technology, Czech Republic
Lenka Turoňová Brno University of Technology, Czech Republic
Additional Reviewers
Kfir Barhum Pavel Čeleda
Hans-Joachim Boeckenhauer Vojtěch Forejt
Yu-Fang Chen Lukáš Holík
Organization IX
Ivan Kolesár Jakub Pawlewicz

Jan Křetínský Martin Plátek
Sacha Krug Fernando Rosa-Velardo
Julio Mariño Václav Šimek
František Mráz Marek Trtík
Mads Chr. Olesen
Contents
An Open-Source Hardware Approach for High Performance Low-Cost

QoS Monitoring of VoIP Traffic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Gianni Antichi, Lisa Donatini, Rosario G. Garroppo, Stefano Giordano,
and Andrew W. Moore
Today’s Challenges for Embedded ASR . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Jozef Ivanecký and Stephan Mehlhase
Automating Software Analysis at Large Scale . . . . . . . . . . . . . . . . . . . . . . 30

Daniel Kroening and Michael Tautschnig
3D Model-Based Segmentation of 3D Biomedical Images . . . . . . . . . . . . . . 40

Stefan Wörz
LTL Model Checking of LLVM Bitcode with Symbolic Data . . . . . . . . . . . 47

Petr Bauch, Vojtěch Havel, and Jiří Barnat
Steiner Tree 1.39-Approximation in Practice . . . . . . . . . . . . . . . . . . . . . . . 60

Stephan Beyer and Markus Chimani
On Monitoring C/C++ Transactional Memory Programs . . . . . . . . . . . . . . . 73

Jan Fiedor, Zdeněk Letko, João Lourenço, and Tomáš Vojnar
Bent Functions Synthesis on Intel Xeon Phi Coprocessor . . . . . . . . . . . . . . 88

Radek Hrbacek
Parallelisation of the 3D Fast Fourier Transform Using the Hybrid

OpenMP/MPI Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Vojtech Nikl and Jiri Jaros
Mapping Problems to Skills Combining Expert Opinion and Student Data . . . . 113
Juraj Nižnan, Radek Pelánek, and Jiří Řihák
Image Analysis of Gene Locus Positions Within Chromosome Territories

in Human Lymphocytes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Karel Štěpka and Martin Falk
Context-Switch-Directed Verification in DIVINE . . . . . . . . . . . . . . . . . . . . 135

Vladimír Štill, Petr Ročkai, and Jiří Barnat
A New Concept in Advice Complexity of Job Shop Scheduling. . . . . . . . . . 147

David Wehner
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

An Open-Source Hardware Approach
for High Performance Low-Cost QoS Monitoring
of VoIP Traffic
Gianni Antichi1(B) , Lisa Donatini2 , Rosario G. Garroppo2 , Stefano Giordano2 ,

and Andrew W. Moore1
1
Computer Laboratory, University of Cambridge, Cambridge, UK
gianni.antichi@cl.cam.ac.uk
2
Department of Information Engineering, University of Pisa, Pisa, Italy
Abstract. A key issue in VoIP services market is the availability of

tools that permit a constant monitoring of the relevant Quality of Ser-
vice (QoS) parameters. Several commercial and open-source solutions
are available, based on dedicated hardware and/or open-source soft-
ware. These solutions aim to achieve a tradeoff between performance
and instrumentation cost. In general, high performance and precise mon-
itoring tools are based on dedicated hardware, which is expensive. In
contrast, cheaper software-based solutions working on top of a Commer-
cially available Off-The-Shelf (COTS) hardware are performance-limited,
especially when serving high-capacity links. In this context, the paper
presents an open-source solution for QoS monitoring of VoIP traffic that
achieves high performance at significantly lower costs. The proposed
solution exploits the performance capabilities achievable with a field-
programmable gate array (FPGA). The associated costs reduction arises
from the high flexibility of an FPGA. Our experimental analysis explores
the accuracy of the developed prototype, measuring against relevant QoS
parameters of VoIP traffic on high capacity links.
Keywords: QoS · High performance · Open-source · FPGA · VoIP
1 Introduction
Building, running and maintaining enterprise networks is getting more compli-
cated and difficult. Part of the problem is the proliferation of real-time appli-
cations (i.e., voice, video, gaming) that demand more and more bandwidth and
low-latency connections. The growth of broadband networks, the continuously
increasing in link speed is driving end-users to an increasing interest towards such
applications. As a result, network management is getting everyday more chal-
lenging for network operators. In particular, voice over Internet Protocol (VoIP)
services have proved to be a disruptive technology that has transformed the
telecommunication industry. In this scenario, providing end-to-end quality rep-
resent a key point for VoIP service providers. Monitoring and improving network

P. Hliněný et al. (Eds.): MEMICS 2014, LNCS 8934, pp. 1–15, 2014.
DOI: 10.1007/978-3-319-14896-0 1
2 G. Antichi et al.
performance parameters like packet loss rate, delay or jitter is the widespread
approach to guarantee a good VoIP quality.
Such parameters can be monitored using either active or passive approaches.
In the first case, results may not represent the actual user experience as the net-
work performances may be corrupted by the test packets sent to estimate the
call quality. On the other hand, passive measurement tests observe user traffic
without increasing the network load. However, such an approach requires probes
able to process packets without loss. While this is a pretty easy task on low
speed links, it becomes more challenging on high capacity links (i.e. 1G, 10G
and more). This leads to a trade-off between the cost and performance of moni-
toring tools: those based upon dedicated hardware and those utilising lower-cost
but potentially lower performance COTS hardware.
In this scenario, we present an FPGA-based open-hardware approach to
enable QoS monitoring of VoIP services. We built our prototype on top of
the NetFPGA [17], an open networking hardware system that enables rapid
development of hardware-accelerated packet processing applications. The open-
source nature of the system allows an easy porting to different FPGA-based
platforms, making such a solution attractive regardless the current limitations
due to the physical hardware infrastructure used. The continuous rising of new
open-hardware platforms [10] [22] with a better hardware infrastructure increases
the potentiality of the proposed solution.
The paper presents the related work in section 2, while section 3 describes
the RTP detection scheme implemented for the proposed framework and reports
the results of its performance analysis. The strategies used for the estimation of
the IP level QoS parameters are summarised in section 4. Details of the architec-
ture of the developed open-source hardware platform are shown in section 5, while
section 6 presents the results of the performance analysis. Section 7 concludes the
paper.
2 Related Work
QoS monitoring of VoIP traffic is not a new field. Commercial tools, such as
[7], [6], [5] and [4] are availbale on the market. They are able to sustain high
traffic rates but on the other hand are closed-propretary and expesinve. There
are VoIP monitoring functions integrated in tools such as [8], but that is a
general tool for packet sniffing and monitoring, its purpose is not to monitor
VoIP continuously. Open source-software based solutions like [13] or [18] are
also available, However, in high speed scenarios (i.e., 1Gbps or higher) solutions
like that require expensive specialised hardware to work properly.
3 The RTP Detection Scheme

The best and easiest way to detect RTP traffic is based on the analysis of
signalling protocols (i.e., SIP, H.323, etc). Unfortunately, signalling and RTP
packets may have different routing, making such approach unusable. Exploiting
An Open-Source Hardware Approach 3
RTCP packets in the detection process can be a solution to the problem, since
RTCP packets have more predictable fields than RTP. Such an approach result
in poor detection performance when the RTCP packet rate is very low compared
to the RTP one. A third way is based on the recognition of statistical traffic pat-
terns (e.g., packet length, inter-packet arrival time). However, such a procedure
need a training phase or/and the knowledge of the statistical properties of RTP
flows to be effective. We propose a two-stage scheme, where the pre-filtering
stage is used to reduce the load of the second stage. Such an algorithm is able to
recognise an RTP flow without the support of any RTCP or signalling packet,
using heuristics that exploit the features of some RTP header fields.
In particular, as shown in Figure 1, the considered RTP detection scheme is
based on two different steps:
– Pre-Filtering : The Deterministic Filtering
– Flow analysis : The Heuristic Filtering
The first stage is used to deterministically pre-filter all the packets that
cannot belong to an RTP stream (e.g., TCP flows), while the second is actually
in charge of detecting RTP streams.
Fig. 1. The proposed RTP detection scheme
3.1 Pre-Filtering: The Deterministic Filtering

Only the packets that match the following rules are passed to the next stage:
– use IP/UDP protocol
– use UDP ports that are allowed for RTP traffic (even ports greater than
1024)
– the content of the supposed RTP header field “Protocol Version” is 2
Packets that do not match these rules, will not be considered by the Heuris-
tic Filtering stage. The aforementioned checks represent a necessary but not
sufficient condition for packets belonging to an RTP stream.
4 G. Antichi et al.
Fig. 2. RTP packet
3.2 Flow Analysis: The Heuristic Filtering
We define as a flow all packets belonging the same 6-tuple (IP destination
address, IP source address, layer 4 destination port, layer 4 source port, RTP
Payload Type, RTP Synchronisation Source Identifier). RTP Payload Type indi-
cates the format of the payload and determines its interpretation by the applica-
tion while RTP Synchronisation Source Identifier (SSRC) uniquely identifies the
source in a session and allow to distinguish different sessions of the same source.
It is worth highlighting that our detection schema separate each direction of a
flow because they can have different QoS parameters. Information taken from
each new arriving packet is compared with data acquired from previous packets
belonging the same flow. An increasing value of the supposed RTP header field
“Sequence Number” (SN) in consecutive packets of a same flow, could indicate
the flow might be RTP (at this stage we are not sure yet it is actually RTP). If
such a pattern is recognised in an adequate (we will define it better afterwards)
number of consecutive packets, the flow can be considered an RTP stream. There
are two main parameters that need to be carefully set in this process:
– Sequence Number Range (SNR): The range of accepted SN values

– Number of observed packets (NoP): The threshold actually used to detect
the RTP stream
The need for a range of accepted SN values comes from the fact that packets
could be lost in the network or suffer from out-of-order arrivals. Accepting only
packets with perfectly consecutive SN values may lead to false negatives. NoP
represents the minimum number of received packets falling in the SNR necessary
to identify a flow as RTP avoiding False Positive (FP).
The pseudocode 1 shows how the heuristic filtering stage works. The counter
matching pkts is incremented every time the SN value of a new arriving packet
matches the proposed range. The stream is recognised as RTP when such a
counter reaches the N oP value.
Algorithm 1. Pseudo-code for the Heuristic Filtering stage. stored sn is the

last received Sequence Number, new sn represents the new received Sequence
Number and matching pkts is the number of packets that match the range
1: if (new sn > stored sn) && (new sn < stored sn + SN R) then
2: matching pkts + +
3: if matching pkts == N oP then
4: RT P DET ECT ED
5: end if
6: end if
7: if new sn > stored sn then
8: stored sn ← new sn
9: end if
new sn replaces stored sn only if it is higher regardless if it falls in SNR.

This is to be able to recognise flows even in case of packet loss or out-of-order
arrivals. It is important to highlight that the system does not cancel or invalidate
an entry if multiple packets with unmatched SN values have been received, in
order to avoid continuous adding and deleting of streams that suffer of out-of-
order packets arrival. Problems and solutions relating to memory management
are discussed in the architecture section (5).
NoP and SNR must be set carefully. Both parameters influence directly the
amount of time the algorithm takes to detect an RTP stream and the False
Positive (FP) ratio. Low SNR and high NoP values prevent FP, but make it
harder to detect RTP streams in presence of out-of-order arrivals or packets
loss. On the other hand, higher SNR and smaller NoP values increase the FP
ratio, but enable a faster RTP detection.
We made calculations to understand which values for SNR and NoP would
minimise the probabilities of False Positives (FP) and False negatives (FN).
F alse P ositives: this is the case where a flow is not RTP, yet it is recognised
as RTP. That is, the variable matching pkts is incremented a number NoP of
times, even though there is not an RTP flow. We assume the 16 bits of the
Sequence Number field can be uniformly distributed in the range [0, 65535], thus
the probability PIN F P of increment in a case of FP can be calculated as:
SN R
PIN F P = (1)
65536
since a range of SNR values will be accepted.
PF P may be obtained considering NoP consecutive packets that cause an
increment of matching pkts variable:
PF P = (PIN F P )N oP (2)
F alse N egatives: In this scenario, the observed consecutive Sequence Num-
ber values are actually correlated, and in an ideal condition they are incremental.
However, it is possible to observe a non-incremental increase, even a decrease,
or an increment that is not in the acceptable range, when in presence of packet
6 G. Antichi et al.
loss or out-of-order arrivals. Such a phenomenon could lead to the result that a
packet, actually belonging to an RTP flow, does not pass the check proposed in
line 1 of algorithm 1. We define PAB as the probability that a packet belonging
to an RTP stream does not produce an increment of the matching pkts variable.
As a consequence, 1 − PAB is the probability to have the matching pkts (when
the packet really belongs to an RTP flow). Assuming each “packet in” event
independent, we can model the system as a sequence of independent Bernoulli
random variables. In this case, the probability PIN P R (i) of the event “to have
an increment of matching pkts at the i-th packet” in the case of RTP flow, can
be calculated using the geometric distribution:
i−1
PIN P R (i) = (1 − PAB )PAB f or i = 1, 2, · · · (3)
whereas the probability PIN P R of having an increment of matching pkts,
can be calculated as:
i=SN
R
i−1
PIN P R = (1 − PAB )PAB (4)
i=1
Assuming PAB = 0.01, we obtain PIN P R = 0.9999 and PIN F P = 3.05·e−05

for SN R = 2, and PIN P R = 0.999999 and PIN F P = 4.58 · e−05 for SN R = 3.
We decided to set SN R = 3. Such value guarantees a high probability (almost
1) that an RTP packet entering in the heuristic stage produces a matching pkts
increment. In this condition, the detection time of a RTP flow is strictly related to
N oP and the packets rate of the RTP flow. To reduce the PF P , we set N oP = 5;
this value leads to PF P = 10−22 . Furthermore, in the worst case of a VoIP RTP
flow with packets transporting 40 ms of speech, the packet rate is 25 pps; thus,
the probability of detecting this RTP flow in the first 125 ms of conversation is
almost 1, whereas the PF P is negligible.
It is worth to highlight that our algorithm differs (from the best of our
knowledge) from other well known algorithms already proposed in literature.
Costeux et al. [11] claim the possibility of detecting RTP streams checking the
PT header field (the values of this field must belong to the set {0 − 34, 96 − 127})
and the correctness of the RTP PDU length. Such a solution can be implemented
without keeping in memory any data, allowing on the fly and stateless filtering.
On the other hand, we exploit the SN header field correlation minimising this
way the FP probability. Guntur [16] proposes a similar approach, based on the
aforementioned 2-steps approach (pre-filtering and detection). However, in the
pre-filtering stage we also use the layer 4 ports and the content of the supposed
RTP header field “Protocol Version” in order to reduce the amount of “false
alarms” sent to the second stage. In addiction, we distinguish different flows
using also the supposed SSRC in order to easily distinguish different calls (calls
at different times) between same hosts.
3.3 Algorithm Validation

The algorithm has been validated though extensive tests campaign. We used
PCAP traces with real traffic to evaluate the detection ability and we estimated
the False Positives (FP) and False Negatives (FN) probabilities as follows
NF P
PF P = (5)
NOP
NF N
PF N = (6)
NOP
where
– NOP : the total number of RTP packets
– NF P : the number of non-RTP packets recognized as RTP
– NF N : the number of RTP packets not recognized as RTP
We used real traffic PCAP traces containing RTP streams to test the FN
probability. First, we analysed them with Wireshark [8] to obtain NOP for
each trace. Then, we compared the obtained results with the ones given by the
proposed scheme. The results of these tests showed that the proposed solution
detected all the RTP streams, and PF N was zero. We used then PCAP traffic
data that did not contain any RTP stream to test the FP probability obtaining
a FP probability of zero. In summary, all the performed tests report that the
100% of the RTP traffic has been recognised without any FN or FP.
4 Data Gathering for QoS Monitoring

Voice quality measurements can be carried out using either subjective or objec-
tive methods. The Mean Opinion Score (MOS), defined in [2], is the most widely
used subjective technique. Subjective are used to evaluate either the Listening
Quality (LQ) (i.e., one-way only) or the Conversational Quality (CQ) (i.e., two-
way, in this case users interaction is involved). In voice communication systems,
MOS is the largest accepted performance metric, since it can be directly linked to
the Perceived Quality of Service (PQoS). The MOS value is obtained by asking
a panel of listeners and speakers to grade the quality of speech under controlled
conditions as set out in the ITU-T standard P.800 [2]. In particular, the MOS
scale consists of Excellent, Good, Fair, Poor, Bad, where the value 5 is asso-
ciated to Excellent, whereas 1 to Bad. Subjective measurement methods have
many drawbacks, since they are time consuming and expensive, lack of repeata-
bility, and cannot be used for long-term or large scale voice quality monitoring
in an operational network infrastructure.
Objective measurement techniques may be classified as intrusive or non-
intrusive. Intrusive methods are more accurate, but normally they are unsuitable
for monitoring real-time traffic, as they need a reference data and the injection of
probing traffic into the network. ITU-T P.862 Perceptual Evaluation of Speech
8 G. Antichi et al.
Quality (PESQ) [3] is the most widely used intrusive measurement method to
evaluate the quality of VoIP applications. It is designed for LQ voice measure-
ment and the prediction of MOS value is based on a comparison between a
degraded speech signal and a reference one. Non-intrusive techniques do not
need reference data and can be used to monitor and/or predict the voice quality
directly from the network and system performance metrics (e.g., packet loss,
delay, jitter and codec). The ITU-T E-Model [1] is a computational model that
can be used to non-intrusively predict the voice quality. It allows the estimation
of the expected speech quality given some parameters as the codec, quantising
distortion or IP level QoS parameters [14]. The IP level QoS parameters mon-
itoring plays a key role in the control of the PQoS. In particular, most of the
techniques used for the monitoring of PQoS rely on the observation of:
– Jitter
– End-to-end delay
– Packet loss
4.1 Jitter
Jitter represents the variation in the inter-arrival time between packets. We
recursively calculate it using the following equations [21]:
|Di,i−1 | − Ji−1
Ji = Ji−1 + (7)
16
Di,i−1 = (Si − Si−1 ) − (Ri − Ri−1 ) (8)
where R is the packet arrival time and S the associated timestamp in the RTP
header. Notice that in order to obtain the jitter in seconds, we have to determine
the value in seconds of the timestamp unit. The RTP timestamp is generated
according to the RTP clock rate that is equal to the number of sampling periods
per seconds. As an example in the cases of G.711 or G.729 codec, the sample
clock is 8000 Hz, hence the timestamp unit is equal to 125 μs.
4.2 End-to-End Delay

This parameter represents the time it takes for a packet to travel from source
to destination. Such latency contributes to the “talk-over effect” experienced
during bad VoIP calls, where people end up talking over each other because
they think the other person has stopped speaking, and increases the problems
associated with the echo effects. Delays higher than 250ms make a conversation
difficult. End-to-end delay is recursively calculated according to the following
equations:
delayi − Ei−1
Ei = Ei−1 + (9)
16
delayi = rx timei − tx timei (10)
rx timei is the arrival time of the ith packet, while tx timei is the time when
the packet was sent. When a packet arrives, only the rx time can be estimated.
The tx time is obtained from the associated RTCP stream through the following
equations:
tx time = N T PRT CP + γi (11)

γi = (Si − sRT CP ) ∗ RT P samplerate (12)
Where N T PRT CP is the NTP timestamp of the first RTCP packet received
(that is: the exact time it was sent, in seconds), and sRT CP is its RTP timestamp.
Thus, γi represents the time between the sending of the first RTCP packet and
the ith RTP packet (in seconds). This procedure requires a clock synchronisation
between the RTCP/RTP sources. The synchronisation can be obtained by means
of NTP servers when an accuracy of few ms can be tolerated in the estimation of
this parameter. Otherwise, advanced strategies for clock synchronisation must
be considered.
4.3 Packet Loss

Packet loss happens when one or more packets fail to reach their destination.
In media streaming applications small packet loss values do not affect the intel-
ligibility of the conversation. The number of lost packets is calculated with the
following equation:
lost pkts = expected pkts − rcvd pkts (13)

Where expected pkts is estimated as last seq num − f irst seq num + 1.
Notice that this solution might give inexact results if packets are out of order.
For this reason, our solution tries to overcome this issue with a flag that signals
if there have been out-of-order packets, which means the value for packet loss
might be inexact.
4.4 Distributed QoS Monitoring Architecture

In a general scenario, the proposed solution can also be used as a probe of a
distributed monitoring framework, as shown in figure 3. No modifications are
needed in the RTP flow detection, since this procedure relies only on RTP fea-
tures. Furthermore, in some cases, such as Voice trunking, where operators gen-
erally know the equipments configuration and routing, the RTP flow detection
scheme can be avoided.
We assume that each device can observe at least one RTCP packet for each
identified RTP flow to measure the end-to-end delay. If RTCP packets cannot
be observed, the end-to-end delay estimation can be still performed through a
cooperation amongst probes. Indeed, in this scenario, each device could generate
RTCP-like packets that contain the information necessary for the estimation of
the end-to-end delay parameter.
10 G. Antichi et al.
Fig. 3. The Distributed Architecture
5 Architecture
We developed our architecture on top of the widely used, open-source, NetFPGA-
1G platform. For each RTP stream, statistics (i.e. jitter, packet loss and
end-to-end delay) are estimated in hardware allowing high precision traffic char-
acterisation.
Fig. 4. The overall architecture of the proposed system
Figure 4 shows a high-level block diagram of the proposed solution. Pack-

ets enter the system through the physical interface and are being processed in
pipeline. Both the RTP detection algorithm and the QoS parameters calculation
are performed in the QoS Monitoring module and, for the sake of minimising
FIFO-induced jitter, the timestamping module is inserted just before the receiv-
ing queues.
Timestamper. An accurate timestamp solution is crucial. Statistics like jitter

(section 4.1) and end-to-end delay (section 4.2) strictly depends on the inter-
nal timestamping module. Packets are being timestamped at the very earliest
possible moment in the system to minimise FIFO-induced jitter. Since Ethernet
preamble can have variable length, we sample the timestamp counter when the
Ethernet start-of-frame delimiter arrives – again, to minimise jitter. A times-
tamp system could be obtained from a 64-bit counter driven by the 125 MHz
free-running system clock, giving one increment every 8 ns (in contrast with the
96 ns inter-frame gap of Gigabit Ethernet). Unfortunately, this implementation
provides no means to correct oscillator frequency drift, bringing to a huge per-
formance degradation over time [9]. Such a problem can be solved by means of
Direct Digital Synthesis (DDS) [20], a technique by which arbitrarily variable
frequencies can be generated using FPGA-friendly, purely synchronous, digital
logic. Therefore, we decide to use the IRCT timestamp module proposed in [9].
Such a module proved to be both accurate and stable.
QoS Monitoring. The QoS Monitoring module detects RTP streams (section 3)
and gathers QoS statistics (section 4). Information related to each RTP stream is
stored in the SRAM memory. The NetFPGA SRAM memory is 4.5 MB, divided
in 512k words of 72 bit each. While half is reserved to the Output Queues, the
remaining is used by our data structures. Each stream requires 72B of memory
space to store flow data (IP addresses and ports pair), RTP/RTCP information
needed either for detection or statistic measurements as well as the calculated
statistics. The maximum amount of RTP streams managed at the same time by
the proposed solution comes from this memory limitation: using half SRAM we
can address approximately 32700 RTP flows at the same time. Moving the Out-
put Queues in DRAM or BRAM (the fastest on-chip memory usable in FPGA
design) would allows us to use the whole SRAM for our data structure, doubling
the maximum number of RTP managed at the same time. We point out that the
provided numbers are just board-specific and do not affect the way the overall
architecture works. Every time a packet of a new flow passes the deterministic
filtering, a new memory block in SRAM is needed. We perform a hash on the
flow data (IP source address and source port) to obtain the SRAM address to
store all the flow-related information. If a collision occurs (the SRAM address is
already in use) we check when the last packet of the recorded flow was received.
If the time passed is greater than τ we can suppose the recorded flow is finished,
and we can overwrite the SRAM memory block with the new flow. Otherwise,
we should perform a new hash function in order to find a new SRAM address
(we perform a maximum of four attempts, after which, the new flow will not
be recorded). The choice of τ parameter is crucial. Small values could lead to
the deletion of an active flow while big values could lead to memory leaks. We
decided to set it to 3 minutes [15]. Again, even in this case we decided not to rely
on the associated RTCP stream (i.e., the RTCP BYE packet indicates the end
of a conversation) as we want to assure the correct system operation without
the presence of RTCP packets.
6 Performance Evaluation
We evaluated the performance of the proposed solution in measuring the afore-
mentioned QoS parameters at different traffic loads. We used traffic traces sniffed
during real VoIP calls. In order to recreate as much as possible a real environ-
ment, we connected two VoIP terminals to Internet in different ways (i.e., wired,
wireless). We placed a traffic sniffer (i.e., Wireshark [8]) between the two termi-
nals in order to get useful traffic traces..
6.1 Jitter and Packet Loss
The obtained PCAP traces have been post-processed in order to estimate the
jitter and packet loss of RTP traffic produced in each direction by the considered
call. Then we replayed the PCAP traces into one of the NetFPGA ports using
an accurate hardware packet generator [12]. We point out the importance of
reproducing the traffic reported in the PCAP trace with high fidelity, in terms
of packets content and inter-departure times. Such a precision, guaranteed by
the NetFPGA packet generator, allows us to fairly evaluate the accuracy of the
statistics estimation of our system. Indeed, by means of this procedure we can
compare the statistics measured on-line by the NetFPGA with the results cal-
culated off-line using the PCAP file. These last results represent the reference
values. Figure 5 summarises the results for about 20 data sets; the circle repre-
sents the jitter value estimated on-line with the NetFPGA, whereas the x the
values calculated off-line on the PCAP file. The figure shows the high accuracy
of the NetFPGA system. Quantitatively, the average of the absolute error of the
NetFPGA solution with respect to the reference values is equal to 0.25%.
20
Our Prototype
Base line values
15
Jitter (ms)
10
0
0 5 10 15 20
Traces
Fig. 5. Jitter measurements
As for the packet loss, the number of lost packets calculated by our sys-
tem was compared to that obtained with wireshark, obtaining exactly the same
values, with a 0% error.
6.2 End-to-End Delay

Testing the end-to-end delay estimation is trickier. We could not use traffic traces
but we had to measure it during a live call. For this reason we used a different
testbed (figure 6). We installed a VoIP client on two different hosts (client A
and B). We connected host A as well as host B to our NetFPGA device, so that
every communication between these hosts had to go through it. We also installed
Netem [19] software on both machines, to add an end-to-end delay set during
the starting of the single trial.
Fig. 6. The proposed testbed
Our Prototype
Netem values
500
400
End-to-End delay (ms)
300
200
100
0
0 5 10 15 20 25 30 35 40
Traces
Fig. 7. End to end delay measurements
We sinchronised the two clients and the NetFPGA system with the same
NTP server, and we started a VoIP call. We compared the on-line end-to-end
delay estimation performed by the NetFPGA with the values set in Netem during
each trial. The results summarised in figure (7) qualitatively show the accuracy
of the on-line estimation technique. From a quantitative perspective, the mean
absolute error is of 3.74%.
7 Conclusions and Future Work

The paper introduces a flexible and cost-effective approach to passive VoIP mon-
itoring. The performance analysis of the proposed architecture proved the ability
of estimating the main QoS parameters within an accuracy given by hardware

timestamp. Such interesting results have been obtained guaranteeing high flexi-
bility and low cost with respect ASIC-based solutions. The open-source nature of
the system allows an easy porting to different FPGA-based platforms, making
such a solution attractive regardless the current limitations due to the physi-
cal hardware infrastructure used. The continuous rising of new open-hardware
platforms (i.e., NetFPGA-10G, NetFPGA Sume) with a better hardware infras-
tructure increases the potentiality of the proposed solution. We are planning to
extend the presented architecture as a probe of a distributed monitoring frame-
work which will allow to guarantee QoS parameters estimation even when RTCP
traffic is not available.
References
1. ITU-T G.107 - The E-Model, a computational model for use in transmission plan-
ning
2. ITU-T P.800.1 - Mean Opinion Score (MOS) terminology
3. ITU-T P.862 - Perceptual evaluation of speech quality (PESQ), an objective
method for end-to-end speech quality assessment of narrowband telephone net-
works and speech codecs
4. OmniPeek Network Analyzer. http://www.wildpackets.com/products/omnipeek
network analyzer
5. VoIP & Network Quality Manager. http://www.solarwinds.com/voip-network-
quality-manager.aspx
6. VoIP Monitor. http://www.voipmonitor.org
7. VoIPFuture. http://www.voipfuture.com
8. Wireshark protocol Analyzer (was Ethereal). http://www.wireshark.org
9. Antichi, G., Giordano, S., Miller, D.J., Moore, A.W.: Enabling open-source high
speed network monitoring on netfpga. In: Network Operations and Management
Symposium (NOMS). IEEE/IFIP (2012)
10. Blott, M., Ellithorpe, J., McKeown, N., Vissers, K., Zeng, H.: FPGA research
design platform fuels network advances. Xilinx Xcell Journal (73) (2010)
11. Costeux, J.L., Guyard, F., Bustos, A.M.: Detection and comparison of rtp
and skype traffic and performance. In: Global Telecommunications Conference
(GLOBECOM). IEEE (2006)
12. Covington, G.A., Gibb, G., Lockwood, J.W., Mckeown, N.: A packet generator on
the netfpga platform. In: Symposium on Field Programmable Custom Computing
Machines (FCCM). IEEE (2009)
13. Deri, L.: Open source voip traffic monitoring. In: System Administration and Net-
work Engineering Conference (SANE) (2006)
14. Garroppo, R.G., Giordano, S., Iacono, D., Tavanti, L.: Experimental and simu-
lation analysis of a wimax system in an emergency marine scenario. Computer
Communications (2011)
15. Guha, S., Daswani, N., Jain, R.: An experimental study of the skype peer-to-peer
voip system. In: International Workshop on Peer-to-Peer Systems (IPTPS) (2006)
16. Guntur, R.: Technique for identifying rtp based traffic in core routing switches. In:
Patent Application Number 20090135834 (2009)
17. Lockwood, J.W., McKeown, N., Watson, G., Gibb, G., Hartke, P., Naous, J.,
Raghuraman, R., Luo, J.: Netfpga-an open platform for gigabit-rate network
switching and routing. In: International Conference on Microelectronic Systems
Education (MSE). IEEE (2007)
18. Manousos, M., Apostolacos, S., Grammatikakis, I., Mexis, D., Kagklis, D.,
Sykas, E.D.: Voice-quality monitoring and control for voip. IEEE Internet Com-
puting, 35–42 (2005)
19. Hemminger, S.: Network emulation with netem. In: Australia’s 6th National Linux
Conference (2005)
20. Saul, P.: Direct digital synthesis. In: Circuits and Systems Tutorials (1996)
21. Schulzrinne, H., Casner, S., Frederick, R., Jacobson, V.: Rtp: A transport protocol
for real-time applications (2003)
22. Zilberman, N., Audzevich, Y., Covington, A., Moore, A.W.: NetFPGA SUME:
Toward Research Commodity 100Gb/s. IEEE Micro (2014)
Today’s Challenges for Embedded ASR
Jozef Ivanecký(B) and Stephan Mehlhase
European Media Laboratory, Schloss–Wolfsbrunnenweg 35,

69118 Heidelberg, Germany
{jozef.ivanecky,stephan.mehlhase}@eml.org
Abstract. Automatic Speech Recognition (ASR) is pervading nowadays

to areas unimaginable a few years ago. This progress was achieved mainly
due to massive changes in the “Smart phones world” and ubiquitous
availability of small, and powerful Linux-based hardware.
Recently, the availability of free ASR systems with acceptable speed
and accuracy performance grew. Together with the changes the mobile
world brought, a developer is now able to include ASR quickly without
detailed knowledge of the underlying technology. What will be the future
of embedded ASR systems in this case?
This talk presents two embedded ASR applications and points out
their advantages over today’s quick solutions. The first one demonstrates
how changes in users behavior allowed to design a usable voice enabled
house control. The second one is an extremely reliable in–car real–time
ASR system which can even use a remote ASR for complex tasks.
1 Introduction
Several years ago to deploy an embedded ASR application, the main challenge
was not ASR itself but the hardware, the ASR system was supposed to run
on. Affordable and fast mini PCs did not exist yet. Software-wise suitable HW
usually required expensive and time consuming porting and testing. Significantly
increased performance of affordable Linux based mini PCs allowed to design and
implement a simple, reliable, inexpensive, and especially today, widely usable
speech recognition systems for many different applications. Pocket-sized com-
puters can today easily outperform normal desktop PCs from some years ago
and unified development environments allows very fast porting to a new target
hardware.
Another limiting factor for many ASR applications was availability. ASR
systems were usually connected with some device like a PC, a car or installed
access point for ASR with microphone and button. For always available ASR
in a building a theoretical option was a set of microphones installed everywhere
but in a real life such a solution is not acceptable.
Changes in the mobile phones world several years ago helped to face this
problem. Most of the middle-class and state-of-the-art mobile devices today are
equipped also with wireless network support. With such a mobile device a user
does not need to access a microphone attached to a PC or some wall. Via wire-
less network the mobile phones can easily access also local ASR systems. So

P. Hliněný et al. (Eds.): MEMICS 2014, LNCS 8934, pp. 16–29, 2014.
DOI: 10.1007/978-3-319-14896-0 2
Today’s Challenges for Embedded ASR 17
mobile phones are acting as a remote microphone. Besides that, a change in user
behavior has occurred: The mobile phones are already accepted and being used
through all age groups [5].
Mobiles phones today are powerful enough to run even system for auto-
matic speech recognition. Unfortunately, the variety of mobile devices prevents
to design a low cost speech recognition software running on all available mobile
phones reliably. Indeed it is much easier to design a simple application which is
just recording the speech signal. Applying a client–server approach, the recorded
speech signal from the mobile device can be send to a recognition server for pro-
cessing. As already pointed out such a “server” can today also be a cheap device
which has a similar size as the mobile phone itself and can be installed wherever
ASR application is needed.
Today, speech recognition on a mobile phone or on a PC is a common feature.
Despite the fact that it is mainly a remote service the latency and the accuracy
is acceptable for given task. Why then use a local ASR system? We will answer
this question in Section 2. In Section 3 and Section 4 we show two examples of
local ASR systems with focus to aspects described in Section 2. In Section 5 a
brief summary is provided.
2 Local Versus Remote ASR Systems

Free speech recognition services are easy to use today. Their integration into not
speech enabled application is usually not very complicated and that is the reason
why they are experimentally used also in applications clearly inappropriate for
them.
There are two main reasons why such a service is not suitable for many
applications with embedded ASR:
1. Free services usually do not have a specific usage domain. It means in the
background is large vocabulary LM based system. Such a system can recog-
nize everything and therefore it can be integrated into any application but
the larger the vocabulary, the lower the accuracy. If we try to use such a
service in an application with very limited vocabulary size (e. g. 100 words),
then the accuracy will be significantly worse in contrast to ASR system with
100 words vocabulary. In case that instead of LM it is possible to use a
grammar the result will be even better.
Another disadvantage of LM against grammar for embedded ASR applica-
tions is the need of some semantic interpreter. Well written grammars with
semantic tags will outperform comparable LM based system not only in case
of accuracy but also in case of latency and resources required (which can be
very important as we will show in the following sections).
Despite the fact that grammar based systems can be faster and more accu-
rate, they have also some drawbacks. To design a simple grammar is a simple
task. To design a good and robust grammar is a complicated and time con-
suming task. Another problem are out of the grammar (OOG) utterances. A
grammar based system covering just a wanted domain will always recognize
18 J. Ivanecký and S. Mehlhase
a valid sentence. Perhaps with very low confidence score but that’s not very
reliable rejection parameter. A low confidence score will even assigned to not
OOG utterance in a very noisy environment (car driving at a high speed,
people talking in the background, . . . ). Solution for OOG in the grammar
based system can be some simple garbage model as showed in Section 4 but
it makes the entire design even more complex.
2. Latency. There are 2 main sources of latency. The first one depends on vocab-
ulary size, the used technology and the service setting. It is latency generated
by the recognition system itself. More accurate or larger vocabulary usually
results in increased latency. For some applications, real time processing is
critical.
The second latency source is caused by communication with a remote service.
Such a latency can vary between 100ms and several seconds. In case we want
to use it while being mobile (e.,g. from within a moving car) it becomes
necessary to handle the case that the service is sometimes unavailable.
We tested the latency for one of the popular freely available ASR services for
commands covered by grammars used in Section 3 and Section 4. The best
latency was about 1.5 second, but 4 seconds latency was nothing special. As
we will show later, such a result is for the demonstrated real time applications
not acceptable. It is necessary to note that for the testing we used a good
Internet connection. On a mobile network in a moving car we expect even
worse numbers.
In the following two sections we show two real applications with a local
embedded ASR system and point out particular aspect of local grammar based
system in contrast to remote LM based ASR.
3 Voice Enabled House Control

In this example we present a user interface for controlling home equipment such
as lights, blinds or heating via speech. The question of how to provide the user
with an easy to accept and easy to use interface/device is in the research area
still going on. Some suggest the TV as a device that is readily available and
accepted by people [4]. But it has the drawback that it is not mobile and it does
not allow for a speech interface, which has emerged as a preferred input method.
[4] stated that such a systems had the following requirements:
– light weighted
– simple and intuitive to use
– adaptable to physical and psychological changes
– offers various input methods like speech and touchscreens
– reliable
Therefore, we propose a speech interface for controlling home devices that

runs on mobile phones. The mobile phone addresses several of the previously
mentioned requirements in that it is light weighted, simple and intuitive to use.
Another random document with
no related content on Scribd:
“Study your spelling, Charlie.”
“Charlie, come up here and stand by my desk.”
And so throughout the year, Miss Marlowe ignored the facts that
ought to have led to a reformation of this little boy’s habits.
CONSTRUCTIVE TREATMENT
When a child shows he has not been given careful teaching relative
to sex hygiene, go to his mother and advise her to take the child to a
physician. Explain the physical as well as the moral and mental help
it may be to the child to have one of two very slight operations
performed, after which, with proper diet and bathing, the boy may
easily forget his wrong habits.
COMMENTS
Children can best be taught at home on matters of sex hygiene.

This is especially true of children in the lower grades. Mothers, as a
rule, gladly respond to a teacher’s or physician’s suggestions for
improving the health of their children.
ILLUSTRATION (FOURTH GRADE)
Miss Morris, a fourth grade teacher, School Nurse

called together the mothers of her pupils Instructs
and asked a trained nurse who lived in the village to address them on
sex hygiene. After the talk, Miss Morris said: “The subject just
discussed is a most important one. I shall be very glad, indeed, to
make reports to any mothers whose children, in my judgment, need
attention relative to this subject, if it is the wish of the mothers here
present for me to do so.”
A vote was taken and the mothers thus expressed their desire to
have such help as the teacher could render. Thereafter she felt
perfectly free to go to them whenever it seemed necessary to discuss
this great subject, so pertinent to a child’s welfare.
CASE 153 (FOURTH GRADE)
Miss Vane saw a note fall upon Mary Pratt’s desk. She said,
“Mary, bring that note to me.”
The child, she knew had not yet read the note. Greatly
embarrassed, Mary looked questioningly at Clyde Mitchel before
starting toward Miss Vane.
Contrary to the courtesy which teachers Improper Notes
admonish pupils to show, Miss Vane stood
up, opened the note and perused it in the presence of the school.
While she was looking at the note, Clyde Mitchel buried his scarlet
face in his book.
“You wrote this note, didn’t you, Clyde?” asked Miss Vane.
Clyde only nodded “Yes,” and burrowed even deeper into his book.
“This is a shameful note,” said Miss Vane. “It contains words that
no child should ever write or speak. You may stay after school,
Clyde.”
The boys waited at the second corner from the school house for
Clyde after school.
In about ten minutes Clyde came running toward them.
“What did she do, Clyde?” they asked.
“Aw, nothing; she just preached a little and gave me a few licks
that wouldn’t hurt a baby.”
“What was in the note, anyway?”
He told them exactly what was in the note, and a loud “Hurray!”
went up from the group of listeners. The subject of conversation
among these boys as they went on down the street was as full of
unclean words and suggestions as the worst boys in the group could
think up.
If you can not deal with sex subjects privately, with pupils in the
lower grades, do not deal with them at all. Miss Vane made a mistake
in reading or referring to the note in the presence of others. In her
efforts to suppress such foul communications she occasioned a talk
upon the unnamable topics by all of her own room and many in other
rooms as well.
COMMENTS
Public punishment of culprits who offend by talking or writing on

sex subjects only occasions more such talk. It is like trying to quench
fire by brandishing a fire-brand which emits live sparks in every
direction, each one of which starts a conflagration.
ILLUSTRATION (THIRD GRADE)
When Sadie Moore picked up a note from Avoid Spreading

the floor and handed it to Miss Dietz, who Harm
taught the third grade, the teacher allowed no one to see her when
she read the note. She said privately to Sadie: “I desire that you say
nothing to any one about that note. That is the best way to help me in
this matter.” She studied the handwriting and note paper and fixed
the blame to a certainty upon Conrad James. She resolved at once to
keep sharp eyes on that boy, unknown to him, and to see that he had
no chance to have unrestricted conservation with other pupils for a
while. She supervised all play periods and thereby assured herself
that no harm should come to any one of her pupils through
association with him.
CASE 154 (SEVENTH GRADE)
Pearl Goodwin’s mother was a widow of Morbid Sex-

ill-repute in the village. The eighth grade consciousness
girls slighted Pearl hourly. They avoided sitting with her whenever
possible; they gave her too wide a space at the blackboard while the
rest of them stood so close together as to crowd their work; she went
sadly to and from school, walking alone, for none of the others would
walk with her.
The teacher, Miss Terman, herself a native of the village,
understood, and made no effort to change the situation.
One day Pearl brought a shameful note to Miss Terman, saying
that she found it on her desk. Miss Terman was shocked and made
public inquiry as to where the note came from. Some of the girls felt
sorry for Pearl and showed it by their attitude toward her. The writer
was not discovered. Every day, thereafter, for a week, Pearl showed a
similar note to Miss Terman, and the mystery grew and with it
sympathy for Pearl. Daily Miss Terman made a speech about the
notes and asked help in finding out the writer.
Finally, in despair, she consulted the superintendent of the school.
When he heard the history of the case he said:
“I believe that Pearl herself is the writer of those notes. Her mind
has been poisoned on the sex subject by taunts. I believe she is the
only one in your room who would write such notes.”
With this thought in mind, Miss Terman sought evidence of Pearl’s
guilt. She was not long in finding the half leaves in Pearl’s tablet from
which the paper for the notes had been torn. She even found Pearl
writing a note, and got her pitiful confession of taking this way to call
attention and sympathy to herself.
Miss Terman sentenced Pearl to isolation for the remainder of the
school year (about two months). She was compelled to take her seat
as soon as she arrived at school in the morning and at noon, to have
a separate recess from the others, and remain in her seat after school
closed until the other children had time to reach their homes.
Miss Terman should have drawn Pearl into the games of the other
girls early in the year. She should have said to the leader among the
girls, in private. “You have it in your hands to make a classmate
happy or miserable. You, yourself, will enjoy school better if no girl is
made sad and lonely. I know that the other girls will follow your lead
and, therefore, I desire that you invite Pearl Goodwin into your
school games and give her an opportunity to know and like good
company.”
COMMENTS
Miss Terman, by allowing the note-writing to be publicly known,

caused an epidemic of undesirable talk in her school. She kept this in
mind daily by her isolation program for Pearl. It is only when all are
concerned in a question of this kind that a public talk should be
made on questions of sex.
ILLUSTRATION (EIGHTH GRADE)
Enoch Fites found the disgraceful Hygienic Toilet

condition of the toilet rooms belonging to Rooms
his school to be a source of great temptation and danger to his
pupils. He first solved the general problem of winning his pupils’
confidence. He was a master in quietly introducing improvements in
the school. For example, he secured funds for a splendid clock, which
was connected with the Western Union Telegraph wires and was
corrected every hour. He established a manual training department
and set every boy in high school and in the eighth grade at a bench.
He opened up a domestic science department. He organized tennis
teams and put through a large number of important measures.
When the appropriate time came, he found no difficulty in putting
the toilet rooms for boys in a sanitary condition and keeping them
so. He remarked to a visitor,
“I have not inspected those rooms for two months, but I know just
how they are kept.”
“How in the world do you manage it?”
“I put it up to my boys. I made the toilet rooms entirely adequate
for their needs and then put it up to my boys to keep them clean.
They have never disappointed me.”
3. Meeting the Boy and Girl Question
CASE 155 (HIGH SCHOOL)

When Mr. Harley went to take up his work as superintendent of
the Jamesville High School, he said to a teacher who had served
there the year before: “I believe in preparedness—what was your
greatest disciplinary problem last year?”
“Parties, without a doubt,” she replied. High School
“The last party or the coming party Parties
occupied the minds of the students to the exclusion of their studies.
They were out late at night and consequently did mediocre work,
even the brightest of them.”
“Was nothing done to stop party going?”
“Well, you see, many of the parents upheld the pupils in what they
called their social education, so Mr. Turner (the former
superintendent) didn’t try to prohibit parties.”
“I’m glad to have this information,” replied Mr. Harley.
Later, when the pupils were known to be planning a hallowe’en
party, Mr. Harley announced that he would suspend every pupil who
attended any party at any time during the school year, without first
securing his permission, and that such permission would be given
only very rarely.
A storm of protest from the pupils was seconded by several
mothers, who called upon Mr. Harley to discuss the social aspect of
education.
When, after a nerve-racking day, he told Mrs. Hines, the leading
society woman of the village, that he must carry out his own plans
unaided by the parents, he unwisely aroused the opposition of so
many of his patrons that his work in Jamesville was very seriously
handicapped and he resigned at the end of his first year there.
A Parent-Teachers’ Club should be organized in every school. Early

in the year a meeting of the club should be devoted to the discussion
of out-of-school-hour entertainments. The superintendent should
have the pros and cons presented before the club by both parents
and teachers. The teacher who upholds parties should advise
mothers to talk often with their children upon the subject of
desirable companions; to forego all teasing of the sons and daughters
about “girls” and “beaux”; and to see to it that the young people have
wise chaperons.
COMMENTS
Much of the mischief that arises from parties is due to parents.

Realizing this to be the case, teachers should find a way to talk to
mothers about how to win and hold the confidence of their children
during the trying high school period. The girls should also be
admonished by their teachers to talk to their mothers freely about
their social affairs.
ILLUSTRATION 1 (HIGH SCHOOL)
Miss Fanson was a high school teacher who was justly admired by
the girls under her care. She had talked to the girls about the
deference and homage which they should show to their parents in
social matters. Alice Grant believed that Miss Fanson was exactly
right, hence was willing to act upon her teacher’s advice.
Since she had entered high school, boys had suddenly become very
interesting to Alice. She blushed one afternoon as she plucked up her
courage to reveal certain developments to her mother.
“Mother, the Freshmen are going to give a party, and a boy in my
German class has asked me to go. May I?” Her voice affected
indifference.
But Mrs. Grant knew her young daughter Retaining
and saw through that coolness. Her Alice Control
was excited and flushed and happy over a boy! And she stared
blankly for a moment as the realization forced its way. Then a
tempestuous refusal from a heart that resented her little girl’s
growing up sprang swiftly to her lips, but she kept back the words. It
did, indeed, hurt to have Alice begin to be a young lady, but could
even she, the most adoring of mothers, restrain time and the youth
that was blossoming in her child?
“I’ll have to think it over, Alice. I’ll tell you in the morning.”
And Alice went to her studying, confident that, whatever her
mother decided, she would be just and allow only big reasons to
weigh with her.
Mrs. Grant thought it over and that night talked it over with her
husband.
“She’s absurdly young—only fifteen,” he objected.
“Yes, but absurdly natural, too, and strong in her desires. I fear, if I
refuse, it may only surround boys with a mysterious glamour for her,
and she might then be tempted to associate with them in spite of me,
and any secrecy or deceit just now is dangerous. And you know our
Alice is growing pretty.”
Mr. Grant regretted and bemoaned the loss of his little girl, but
agreed. “But who is this boy?” he demanded. “Do you know him?”
“No. But I’m going to know all her friends from now on.”
And next morning, when Alice, pink-cheeked and eager-eyed,
sought her mother’s decision, she welcomed the “Yes” with a little
squeal of delight.
“But I’ve been thinking, Alice,” her mother added, “that I’d like to
know the boys and girls you’re going with. Wouldn’t you like to ask
some of them over here some evening before the party?”
“Would I? Well, rather! Mother, you’re a dear.”
“And what about a dress. I suppose you’d like a new one?” Further
question was stifled by an enthusiastic hug.
So they talked of the party and the dress, and then it was not far to
“the boys” and Alice’s new feeling for them. And Mrs. Grant felt that
the sweet intimacy she was entering with this new daughter more
than compensated for the loss of the little girl, who had suddenly
become a young woman.
When Alice returned from the party her mother showed interest in
each detail that her daughter related. She remarked: “You must have
had loads of fun—what did you have to eat? What did you especially
like in the conduct of your classmates?” It is while such concrete
subjects are being discussed that much guidance can be given the
daughter in her formation of opinions as to what is proper or
improper conduct. A teacher who brings about such intimacy as this
incident illustrates has done much for both mother and daughter.
ILLUSTRATION 2 (HIGH SCHOOL)
Miss Canfield took hold of her work with genuine interest as

science teacher in the James Fisk High School. Her knowledge of girl
nature was sufficient to save her from many blunders. Mary Turner
was her problem. A giddy set was overturning nearly all of the
constructive work done for her by her teachers.
Miss Canfield decided to go over matters with Mrs. Turner, Mary’s
mother. In the conversation, Mrs. Turner saw where she must take a
hand in Mary’s affairs.
There was no doubt but what Cecily Overcoming
Gregg, a classmate, was having a bad Undesirable
influence on Mary. Mrs. Turner rocked Influences
fitfully between stitches and remembered how sweet and natural
Mary had been before she got so intimate with Cecily. But now she
was catching some of that young lady’s affected ways, and, Mrs.
Turner feared, some of her lack of modesty with boy companions.
Cecily was seventeen, and Mary, a year younger, respected her
opinions greatly, and gloated over her popularity with certain
overdressed and rather sporty youths who took her about to picture
shows and ice cream parlors. Cecily was slowly convincing Mary that
theirs was the type to admire.
And Mrs. Turner had unwittingly let Mary drift so far from her
influence of late, that she felt helpless. She dared say nothing openly
against Cecily. Mary would only flare up in defense and stand more
staunchly for her friend. If she laid down rules, Mary might secretly
break them, and if she tried to make subtle suggestions, the girl was
certain to pounce on her meaning and resent it.
Mary came home from school that day full of plans for her
birthday party.
“Cecily says I must get some new dance records for the victrola.
Ours are all passé. And I’m going to make little crepe paper favors, by
a cute pattern that Cecily knows. And she wants me to ask Cousin
Ralph. Do you think he’d think us too young for him, since he’s
finished college? I’m crazy to have him meet Cecily! He’ll be ‘dippy’
about her.”
While Mary chattered, a thought lodged by Miss Canfield came to
Mrs. Turner. If she couldn’t influence her daughter herself, unaided,
she must reach her through others.
She answered: “Why, I think it would be lovely to ask him, and I’m
sure he’d like to come.”
And so Mary wrote a cordial invitation to Cousin Ralph and her
mother quietly added a postscript that night—a postscript that grew
into an epistle as she told her nephew, a clean-souled and manly
young fellow, of her problem about Mary.
“Can you help me?” she wrote. “A word from you would weigh
much with her. You’re her ideal of young manhood. Let her see that
you are not fascinated by Cecily; she believes her irresistible. Say no
more than you can judge by seeing her at the party, though. That will
be enough.”
His answer to Mary, his “sweet little cousin,” was frank and warm.
His answer to Mrs. Turner was earnest and sympathetic. He would
try.
The great evening came, and with it a gay and brightly dressed
bevy of Mary’s friends. Some were rollicking; some were bashful; but
Mrs. Turner fancied she saw the Cecily stamp on all of them. On all
except Evelyn Lewis, a simple, attractive girl with fine manners. If
Mary would only prefer her to Cecily!
Cousin Ralph arrived late and created a sensation, for he was tall
and good-looking and possessed of polish and charm. He led all the
fun after that and Mrs. Turner saw Mary’s eyes sparkling with pride
in him.
At a late hour the guests took their leave. But Ralph, lingering after
the others had left, talked over the party with Mary and her mother,
for the former was too excited to want the evening to end.
“How did you like the girls?” Mary inquired, eagerly. And just then
Mrs. Turner found an excuse to leave the room.
“Very much, little cousin. They’re a jolly lot of youngsters. And I’m
quite struck with one of them.”
“Oh, I knew you would be. Cecily, of course!”
“Cecily! O, no!” His emphasis was expressive.
“Not Cecily?” Mary was bewildered.
“That would-be chorus girl with come-hither eyes?” he demanded,
and then, seeing her stricken face, added hurriedly, “But maybe she’s
a special friend of yours.”
“Oh, no,—that is,—not so very. But she’s awfully popular.”
“With only one kind of boys, then, and that’s not the sort I’d like to
see you running round with, cousin mine. The girl that took my eye
was—her name was—Evelyn. She’s a peach. Ask me over some time
again when she’s here, will you?”
Mary nodded a little uncertainly, and then promised.
“Mother,” she said, wonderingly, as Mrs. Turner entered the room,
“Ralph likes Evelyn. And she certainly did look pretty tonight. I’m—
crazy about her myself!”
And as Mrs. Turner squeezed her nephew’s hand, she felt
somehow that a new name was about to be substituted for “Cecily” in
Mary’s vocabulary.
Miss Canfield listened attentively to the mother’s report of the
party and of Mary’s drift into better companionship and naturally
lent aid to the scheme in a dozen little ways—assignments of team
work, comments to Mary on certain lovely qualities in Evelyn and
her type of girl, recommendation of books and magazine articles, etc.
Mother and teacher accomplished an important piece of work by
this campaign in which they substituted, in the unformed mind of a
school girl, a correct model of young womanhood in place of a
degraded type.
4. Falling in Love with the Teacher
When pupils fall in love with their teachers, the problem is not
nearly so serious as the same event would be out of school, for the
reason that every normal tradition of school relations is against such
a state of affairs. The teacher stands, as is said so often, in loco
parentis; and if teachers are fit to bear this relation to their pupils,
they can, and will, easily handle any tendencies toward too intimate
relations with their pupils. The treatment for a pupil who develops
too ardent an admiration for a teacher is based upon the process of
de-personalizing the relations between them; for almost always it
will be found that when pupils have fallen in love with their teachers,
it is because, purposely or unconsciously, the relations have been too
personal.
There are two typical cases—that of young girls who fall in love
with an attractive young man teacher, whom usually they hope to
captivate and marry; and that of boys, relatively less mature, who
rarely reach the ridiculousness of such plans, but shower such
attentions as they may upon the object of their affections, and go to
any length to please her. Most young women teachers have the tact
and good sense to manage such cases wisely, keeping the boys within
the bounds of a normal and fairly platonic regard, and often using
their power to bring about the development of a fine idealism and
many manly virtues in their admirers. But the vain young woman
who likes this kind of popularity is not unknown in schools; she is a
nuisance, doing more harm by her vanity than a dozen sensible
colleagues can undo through every means known to good pedagogy.
The teacher is to blame, as a rule, when either of these conditions
develops. Being older and more experienced, he has the upper hand
and can cure the malady, if he will, especially as he has every sane
tradition on his side. The elimination of the dangerous personal
attitude, of opportunities for the expression of regard, of the
personal appeal, and of subtle suggestions of a sentimental nature,
are all in the power of the teacher. It is just a question of whether he
cares to exercise his will and his ingenuity in the interest of a healthy
relation, or whether he chooses rather to have his vanity flattered by
attentions and popularity.
CASE 156 (HIGH SCHOOL)
Annabel Kingsley was an English teacher Appeals to Vanity

in a small, prosperous town. She was a tiny,
sharp-faced girl of about twenty-five, keenly intelligent, clever and
selfseeking. She dressed well; she sought social opportunities; she
made the most of her friends. Before she had been teaching a month
she had won the devoted admiration of all the boys and most of the
girls in her classes and by Thanksgiving the other women teachers
would hardly speak to her, regarding her with that silent scorn which
intelligent women have for their sisters who will not play fair. The
superintendent was divided between amusement and contempt.
Miss Sperry, the mathematics teacher, went to Miss Bulwer, who
had had the Latin and German for years, and had a talk with her.
“My boys and girls come day after day with their algebra only half
learned,” she complained. “They say they don’t have time for it, and
they are losing all their interest, too. But they write great long
compositions for Miss Kingsley that must take hours to do, and now
she talks of getting up a play to be given at Christmas. She seems to
have captured them completely. How does she do it?”
“When you’ve seen as much of teachers as I have you’ll know,”
Miss Bulwer replied, grimly. “I haven’t heard her talk to them, but I
can tell you just how she goes about it. She makes every one of them
think he’s the budding genius of the century. She has Verne Gibbs
reading Ibsen and planning to write a tragedy. I’ll be bound! She has
persuaded Morris Talbot that he can write short stories. Warren
Hughes thinks he’s very remarkable because she told him he could
appreciate Francis Thompson. Maybe he can, but he can also
appreciate Cicero when he’s given half a chance. Every one of those
youngsters thinks that at last he has found a teacher who really sees
what is in him, the great promise to which the rest of us are blind.
Then he proceeds to fall in love with Miss Kingsley to show her that
her interest is not lost. She appeals to the adolescent vanity that they
all have so much of, and she’s making them so insufferably self-
conscious and sentimental and onesided that you and I can’t do
anything with them.”
Miss Sperry watched Miss Kingsley. She saw that the boys who
could use their father’s automobiles vied with each other for the
honor of taking her home on Friday nights—she lived in a
neighboring town; that they hung over her desk before and after
sessions, engaged in interminable discussions of the value of poetry
or the madness of Hamlet. On her birthday her desk was banked
with roses; Miss Sperry wondered how they found out when her
birthday came. Miss Kingsley’s work went very well, but she robbed
every other teacher of the time and energy that fairly belonged to the
other subjects taught. The result was that the poor work caused by
her selfish policy showed in the classes of other teachers. In her own
there was a constant and lively interest, fanned continuously by the
numberless “conferences” with which she kept her hold on her
students. The school was badly demoralized by Christmas, and yet
the real cause of all the trouble appeared to be the one brilliantly
successful teacher on the force.
The principal of a school should see to it that each teacher and

each subject has a fair share of the attention of the students. In this
case, the principal should say to Miss Kingsley, “I notice that a
number of our boys are falling behind with their mathematics, and
Miss Bulwer tells me that Howard Grimes failed in Latin last month
—something never heard of before. I have been looking for the cause,
and I find that most of those who are failing are spending more time
on their English than is fair. You are stimulating them by a personal
appeal to put time on English which really belongs to other studies.
So I am asking you to discontinue your private conferences for the
present; and, moreover, it is not dignified for you to accept
attentions from the boys as though they were your own age; it will
lead to criticism which will hurt your work and your influence.”
Private talks to the boys and girls about their work, following this
restriction of the English teacher’s demands, may help to bring
results. The other teachers should be encouraged to make their work
as appealing as possible, and to show a personal interest in the
bringing up of grades in the neglected studies. Most important of all,
wholesome social conditions may be stimulated by a series of parties
among the high school students, in which normal relations amongst
themselves are encouraged. Such regulations for study as are needed
to keep the boys from too much contact with Miss Kingsley are to be
adopted, without making their object obvious to the pupils.
COMMENTS
The amative impulses of youth are not vicious, but need direction
and control. Self-control, above all else, is to be taught, and the
teaching must often be reinforced by wise, friendly restraint. Frank
friendships are to be encouraged; sickly, silly sentimentality laughed
out of court. If a teacher, instead of standing ready to give this help
and guidance when it is needed, encourages a sentimental devotion,
as Miss Kingsley did, the most fundamental safeguard of youth is
sacrificed—the ideal of controlled emotion, of a conscious saving of a
sacred experience for the future. A large range of interests, a healthy
balance of activities, and a wholesome unconsciousness of self, tend
to keep young people simple and child-like in their emotional lives.
Above all, no teacher has any business to give the impression that he
alone appreciates youth and its promise, or to make his relations
with impressionable boys and girls unduly personal.
ILLUSTRATION (HIGH SCHOOL)
Clarence Miller was an exceedingly “A Wet Blanket”

handsome young teacher in a small village for Infatuation
high school. In his second year of service, Carolyn Brush, daughter of
the great man of the town, decided that she would not return to the
fashionable boarding school which she had been attending, but
would go to the village school and subjugate Clarence Miller, whom
she met during the Christmas vacation. She was very pretty and very
clever, and her stay in a girl’s boarding school had not made her less
romantic than other girls are.
The lessons were easy for her, and during the first few days she
recited brilliantly, hoping to win special attention from the young
principal. He accepted her most studied efforts with the same
pleasant courtesy he gave to all, and then Carolyn tried another plan.
She failed to recite altogether, looking at Mr. Miller with a pitiful,
hurt look whenever he called upon her, and shaking her pretty head
sadly. The village boys and girls, somewhat awed at best by Carolyn’s
pretty clothes and polished manners, and keenly conscious of
everything she did, observed all this with much interest. Carolyn
became more and more enamored of Mr. Miller the more she saw of
him.
One morning she stepped to the desk when there were no other
pupils near. “Mr. Miller,” she said, “I wonder if I may speak to you—
alone—some time? Tonight, after school, perhaps? Just for a
moment. I am in such trouble.”
“Of course you may, Miss Carolyn,” said Mr. Miller, heartily. “I’ll
be glad to help you if I can.”
But Carolyn was not at school that afternoon. She called up the
school by telephone at five after four, however, said that her mother
had required her help that afternoon, and added that they all wanted
Mr. Miller to come up for supper. “And I hope you will, for I do feel
that you can help me. We can talk after supper.”
“Sorry, but I have some work that is going to take my whole
evening, Miss Carolyn. You can tell me about that matter at recess
tomorrow. Please thank your mother for the invitation, and tell her
how sorry I am I can’t come.”
At recess the next morning, Carolyn said, when she was sure no
prying boy lingered near:
“Oh, Mr. Miller, I have been so worried lately I just couldn’t study.
I have a dear friend at school, whom I’ve trusted and loved for two
years more than anyone else. And now I find that she has deceived
me, and it almost breaks my heart. It seems as if everything has just
stopped, you know; life isn’t the same. What can one do? If one can’t
trust one’s friends, what is there one can count upon?” She looked up
at him with tears in her eyes, the lovely picture of disillusioned youth
in its most appealing form. “I just had to talk to some one about it,
and you’re the only person here who is—you know—like myself—who
would understand.”
Mr. Miller neither fell into this fair trap nor shied at it. He said,
“Now, I’ll tell you just what I would do if I were you. You talk to your
father. He knows all about people, and he’ll give you more good
advice in a minute than I could in a year. If it were I, and a girl had
treated me like that, I’d find a better chum and let her go, and not
weep over it either. Just stop worrying about her. You can’t afford to
lose out on your lessons for a snip of a girl who doesn’t know a good
friend when she has one. Oh—you’ll excuse me, won’t you? I
promised the boys to show them a new curve, and here they are for
me.” And the cautious, sensible principal vanished out-of-doors.
Carolyn, being really infatuated, made one more attempt. “I know
you don’t like me,” she told the principal one day. “But why is it?
What have I done, that you should hate me so? I have tried to get my
lessons, and tried to be good in school; but you seem to hate the very
sight of me.”
“Now, that’s all nonsense,” Mr. Miller averred. “I like you just as
well as anyone else in the room, and, so far as I know and intend, I
treat you just as I do the others.”
To be treated just as the others were treated, was exactly what
Carolyn did not want. She suddenly discovered that the principal was
not handsome, and that she did not care for him. She told her father
that the school was so much poorer than Grey Gables that she
wanted to go back there, and at the Easter vacation she left the high
school. So Carolyn came and went, and not one of the other pupils
knew of the little comedy of sentiment and sense that had taken
place there that winter.
The quickly-veering emotion of youth is easily stimulated or
inhibited by suggestion. Mr. Miller saw through the schemes of his
pupil, and, instead of falling in with them, as he might have done had
he wanted excitement or adulation or romantic adventure, he cut
them off in a friendly, but matter-of-fact way that nipped expectation
in the bud. A flirtation with his pretty pupil might have been a great
deal of fun, but it would have marred his influence with the people of
the village and with his pupils; and he was wise enough to deny
himself that fun for the sake of his professional duty. He might have
stimulated an adventure in half a dozen ways; he steadily declined
even to suggest the thought of such a thing until Carolyn was cured
of her fancy. Without humiliating her in any way, before the other
students, he kept his relations with her impersonal and free of
romantic elements, and so gradually overcame her infatuation by
giving it nothing to feed upon.
DIVISION X
Who breaks his faith no faith is held with him.

—Cervantes
Always act in such a way as to secure the love of your neighbor.

—Cato
AN ILLUSTRATIVE CONTRAST BETWEEN
FAILURE AND SUCCESS
Mr. Bradley was principal for two years of the Newcastle school.
He revealed his characteristics as a teacher so fully that we find in
him an example of the type not to be recommended and yet one that
is very instructive for students of school discipline.
In stature he was slightly below medium height. He came from
rural ancestry and was fairly well equipped as to physique. He had
black hair and eyes, somewhat mobile features and a wandering
gaze. His movements could hardly be called quick, but they were
prompt and without distinct mannerisms.
He had a most gracious manner when meeting people on the street
or in their homes. He spoke kindly to everyone and had the
reputation among the townspeople of being a royal, good fellow.
Even his pupils could not deny that he treated them very courteously
and jovially outside of school hours.
Despite all this he used essentially the method of the hen-pecking
incompetent when handling disciplinary matters in school. The
moment he entered the school precincts he was a different man. His
countenance then betrayed the sternness of the schoolmaster who
dwelt within and apart from the polite gentleman he seemed to be
when outside the school-room. His eyebrows gathered and his
muscles reverberated with the sense of authority that flooded his
whole nature.
His eye was on the lookout for misdemeanors and if a pupil made a
misstep in the realm where Mr. Bradley thought he had jurisdiction,
that harsh, strident voice, with but the slightest trace of fellow-
feeling, spoke the word of correction or announced an impending
penalty.
In the school-room it was his delight to slip up behind an offender
and pluck him by the ear as a reminder of duty. Being the only
instructor who indulged in this practice it soon came to be one of the
most odious signals of his presence in the room. When absorbed in
his subject he made instruction interesting; his pupils could not fail
to learn if they did not venture to vary the program by misconduct.
However, their recollection of his general attitude toward them, the
ease with which they could upset his plans by introducing a few
school pranks, the certainty that he would lose his temper on slight
provocation, always hung as a barrage screen between them and
undivided concentration on the subject-matter of their lessons.
Mr. Bradley made it a practice to watch for accumulating offenses.
He felt incompetent to handle minor evils, but attempted to squelch
a wayward pupil by reciting a list of grievances and applying
penalties for the same. He had a good memory for facts of this sort.
He could shake his finger in the face of a boy or girl and say, “Didn’t
you pull Esther’s hair yesterday ... trip up Jimmie on the way to class
in geometry and purposely spill the crayons when you were at the
board? Now, I have had enough of this. I want to know what you are
going to do about it.”
This gentleman could not catch the drift of things. Early in his first
year Mr. Bradley’s attention rested upon Ted. Ted was a short,
heavy-set chap of some fourteen years, incapable of any
revolutionary propensities, but able to interest himself with a variety
of aggravating tricks. His pranks were individually almost too small
to command severe penalties, but they were too annoying to escape
the principal’s eye.
Unfortunately, Mr. Bradley hit upon the lash as a cure for Ted.
Selecting a more pronounced misdemeanor as an opportunity for
settling accounts with the troublesome pupil, he gave him a sound
whipping.
There was some ground for the general protest that arose from the
high school. Ted was a favorite with every one. The crude principal
had struck one but he had wounded all. His untactfulness had made
him abhorrent to all, even to those who had not hitherto drawn upon
themselves his specific disapproval and useless punishments. Mr.
Bradley, perhaps, never knew that he had undermined his own
usefulness as much by this treatment of a school favorite as by any
single deed that transpired during his whole stay in Newcastle.
He had his own method of handling the problem of whispering. He
made it a rule that every pupil in high school must answer at roll call
at the end of the day on the matter of whispering. If a pupil had
whispered he must answer “Present,” and specify the number of
times during the day he had whispered. If he had a clear record on

Download textbook Mathematical And Engineering Methods In Computer Science 9Th International Doctoral Workshop Memics 2014 Telc Czech Republic October 17 19 2014 Revised Selected Papers 1St Edition Petr Hlineny ebook all chapter pdf

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Download textbook Mathematical And Engineering Methods In Computer Science 9Th International Doctoral Workshop Memics 2014 Telc Czech Republic October 17 19 2014 Revised Selected Papers 1St Edition Petr Hlineny ebook all chapter pdf

Uploaded by

Copyright:

Available Formats

Mathematical and Engineering Methods

in Computer Science 9th International

Combinatorial Algorithms 25th International Workshop

Information Search Integration and Personalization 9th

Semantic Web Collaborative Spaces Second International

Medical Computer Vision Algorithms for Big Data

Biometric Authentication First International Workshop

Parameterized and Exact Computation 9th International

Zdeněk Dvořák Petr Matula

Jiří Jaroš Karel Pala

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Library of Congress Control Number: 2014960061

LNCS Sublibrary: SL2 – Programming and Software Engineering

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

December 2014 Petr Hliněný

Program Committee Track Co-chairs

Hana Kubátová Czech Technical University in Prague,

Ivan Kolesár Jakub Pawlewicz

An Open-Source Hardware Approach for High Performance Low-Cost

Today’s Challenges for Embedded ASR . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Automating Software Analysis at Large Scale . . . . . . . . . . . . . . . . . . . . . . 30

3D Model-Based Segmentation of 3D Biomedical Images . . . . . . . . . . . . . . 40

LTL Model Checking of LLVM Bitcode with Symbolic Data . . . . . . . . . . . 47

Steiner Tree 1.39-Approximation in Practice . . . . . . . . . . . . . . . . . . . . . . . 60

On Monitoring C/C++ Transactional Memory Programs . . . . . . . . . . . . . . . 73

Bent Functions Synthesis on Intel Xeon Phi Coprocessor . . . . . . . . . . . . . . 88

Parallelisation of the 3D Fast Fourier Transform Using the Hybrid

Image Analysis of Gene Locus Positions Within Chromosome Territories

Context-Switch-Directed Verification in DIVINE . . . . . . . . . . . . . . . . . . . . 135

A New Concept in Advice Complexity of Job Shop Scheduling. . . . . . . . . . 147

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

Gianni Antichi1(B) , Lisa Donatini2 , Rosario G. Garroppo2 , Stefano Giordano2 ,

Abstract. A key issue in VoIP services market is the availability of

Keywords: QoS · High performance · Open-source · FPGA · VoIP

3 The RTP Detection Scheme

Fig. 1. The proposed RTP detection scheme

3.1 Pre-Filtering: The Deterministic Filtering

Fig. 2. RTP packet

3.2 Flow Analysis: The Heuristic Filtering

– Sequence Number Range (SNR): The range of accepted SN values

Algorithm 1. Pseudo-code for the Heuristic Filtering stage. stored sn is the

new sn replaces stored sn only if it is higher regardless if it falls in SNR.

Assuming PAB = 0.01, we obtain PIN P R = 0.9999 and PIN F P = 3.05·e−05

3.3 Algorithm Validation

4 Data Gathering for QoS Monitoring

4.2 End-to-End Delay

tx time = N T PRT CP + γi (11)

4.3 Packet Loss

lost pkts = expected pkts − rcvd pkts (13)

4.4 Distributed QoS Monitoring Architecture

Fig. 3. The Distributed Architecture

Fig. 4. The overall architecture of the proposed system

Figure 4 shows a high-level block diagram of the proposed solution. Pack-

Timestamper. An accurate timestamp solution is crucial. Statistics like jitter

6.1 Jitter and Packet Loss

Fig. 5. Jitter measurements

6.2 End-to-End Delay

Fig. 6. The proposed testbed

Fig. 7. End to end delay measurements

7 Conclusions and Future Work

of estimating the main QoS parameters within an accuracy given by hardware

Jozef Ivanecký(B) and Stephan Mehlhase

European Media Laboratory, Schloss–Wolfsbrunnenweg 35,