Professional Documents
Culture Documents
Project Report
<Version 1.0>
Industrial Training
December,2021
Place:
Date:
Sign:
DECLARATION
We hereby declare that this Project Report title Technobot Using Python
submitted by us and approved by our project guide Mr. Anurag Gupta , College of
Computing Sciences and Information Technology (CCSIT), Teerthanker Mahaveer
University, Moradabad, is a bonafide work undertaken by us and it is not
submitted to any other University or Institution for the award of any degree
diploma / certificate or published any time before.
Project Group
Project Guide
(external)
Table of Contents
TMU-FOE&CS Version 5.0 T004A-Project Report
1 PROJECT TITLE 6
2 PROBLEM STATEMENT 6
3 PROJECT DESCRIPTION 6
3.1 SCOPE OF THE WORK 6
3.2 PROJECT MODULES 6
3.3 CONTEXT DIAGRAM (HIGH LEVEL) 6
4 IMPLEMENTATION METHODOLOGY 6
5 TECHNOLOGIES TO BE USED 6
5.1 SOFTWARE PLATFORM 6
5.2 HARDWARE PLATFORM 7
5.3 TOOLS, IF ANY 7
6 ADVANTAGES OF THIS PROJECT 7
7 ASSUMPTIONS, IF ANY 7
8 FUTURE SCOPE AND FURTHER ENHANCEMENT OF THE PROJECT 7
9 PROJECT REPOSITORY LOCATION 7
10 DEFINITIONS, ACRONYMS, AND ABBREVIATIONS 8
11 CONCLUSION 8
12 REFERENCES 9
Appendix
A: Data Flow Diagram (DFD)
B: Entity Relationship Diagram (ERD)
C: Use Case Diagram (UCD)
D: Data Dictionary (DD)
E: Screen Shots
Project Title
In present era , it is believe that Integration is key to all technologies , but issue is
that every user doesn’t have knowledge of all technologies , so we are making a
technical assistant "TECHNOBOT", which will take care of industrial needs in any
technical domains.
Problem Statement
Initially , Manual testing and software were ok , but nowadays world is moving
forward towards automation. Every company wants that there software should
be automated and can able to control multiple different technologies . Its very
typical task to learn every technology deeply and to remember their commands.
And also many peoples don’t have knowledge in these domains so this will help
them also.
Project Description
Implementation Methodology
2 Technologies to be used
2.1 Software Platform
a) Front-end - Web-App
b) Back-end - Python
3 Assumptions, if any
<Guidelines: Mention NONE, if there are NO Assumptions>
Conclusion
We can conclude that if any company use our Software , it will save money and
time for them as everything would be automated so they dont need to hier a
particular technology specialist for thier project.
References
LITERATURE REVIEW
The page describes the key features of the automatic speech recognition systems
training process that can be used in a work of chat bot or voice assistant.
reviewed. Kaldi and CMU. Sphinx has been used as an open source systems’
example. The whole installation and training process are reviewed, instruction for
further researchers are documented. Examined systems have been trained with
different parameter sets and training data. As a result, the most important
training parameters of automatic speech recognition system for chat bot are
designated and described. Results of this research can help to significantly
improve the speed and quality of the speech recognition in chat bots and voice
assistants.
In this paper, the design and development of an intelligent voice recognition chat
bot. The paper presents a technology demonstrator to verify a proposed
framework required to support such a bot (a Web service). While a black box
approach is used, by controlling the communication structure, to and from the
Web-service, the Web-service allows all types of clients to communicate to the
server from any platform. The service provided is accessible through a generated
interface which allows for seamless XML processing; whereby the extensibility
improves the lifespan of such a service. By introducing an artificial brain, the Web-
based bot generates customized user responses, aligned to the desired character.
Questions asked to the bot, which is not understood is further processed using a
third-party expert system (an online intelligent research assistant), and the
response is archived, improving the artificial brain capabilities for future
generation of responses.
In this paper, explains the external language models (LM) integration remains a
challenging task for end-to-end (E2E) automatic speech recognition (ASR) which
has no clear division between acoustic and language models. In this work, we
propose an internal LM estimation (ILME) method to facilitate a more effective
integration of the external LM with all pre-existing E2E models with no additional
model training, including the most popular recurrent neural network transducer
(RNN-T) and attention-based encoder-decoder (AED) models. Trained with audio-
transcript pairs, an E2E model implicitly learns an internal LM that characterizes
the training data in the source domain. With ILME, the internal LM scores of an
E2E model are estimated and subtracted from the log-linear interpolation
between the scores of the E2E model and the external LM. The internal LM scores
are approximated as the output of an E2E model when eliminating its acoustic
components. ILME can alleviate the domain mismatch between training and
testing, or improve the multi-domain E2E ASR.
In this paper, it explains that over the past few decades, there has been
tremendous development in machine learning paradigms used in automatic
speech recognition (ASR) for home automation to space exploration. Though
commercial speech recognizers are available for certain well-defined applications
like dictation and transcription, many issues in ASR like recognition in noisy
research and development of automatic speech recognition during the last few
decades so as to provide a technological perspective and an appreciation of the
fundamental progress that has been made in this important area of information
and communication technology.
In this paper, explains the Frame segmentation which is dividing a signal into
frames, is one of the most important part in speech recognition. In general, the
frame’s size of all segmented frames are fixed at around 20–40 ms, which leads to
the occurrence of a non-periodic signal in each frame. Consequently, the spectral
leakage arise after Discrete Fourier transform on these non-periodic frames in the
process of MFCC and effects on reducing speech recognition performance. In this
paper, pitch segmentation is proposed for reducing the spectral leakage issue by
applied a new technique of pitch detection to produce a periodic signal in each
frame. The neural network models are trained from collected speech signals as
well as the result of performance between segmentations of fixed frame and
pitch frame are compared. As the result, pitch frame gives a higher accuracy than
fixed frame in speech recognition.
the 99.89% accuracy most among all Similarly, Double Q learning algorithm also
given high accuracy but in MATLAB dataset only.
In this work, a new challenging digit Arabic dataset is collected from different
study levels of schools. A large dataset is collected after paying vast effort for
distributing and collecting digit forms over hundreds of primary, high, college
students. After we find that there are few and not challenging Arabic digit
dataset, we paid vast effort for preparing such a challenging dataset. Also the
collected dataset is trained using an efficient model of CNN which represents the
current state-of-the-art for variety of applications. Thus we extensively analyzed
the model by carefully selecting their parameters and showing its robustness for
handling our dataset.
In this paper, the Handwritten Digit Recognition using Deep learning methods has
been implemented. The most widely used Machine learning algorithms, KNN,
SVM, RFC and CNN have been trained and tested on the same data in order
acquire the comparison between the classifiers. Utilising these deep learning
techniques, a high amount of accuracy can be obtained. Compared to other
research methods, this method focuses on which classifier works better by
improving the accuracy of classification models by more than 99%. Using Keras as
backend and Tensorflow as the software, a CNN model is able to give accuracy of
about 98.72%. In this initial experiment, CNN gives an accuracy of 98.72%, while
KNN gives an accuracy of 96.67%, while RFC and SVM are not that outstanding.
In this paper, Handwritten digit recognition has immense applications in the field
of medical, banking, student management, and taxation process etc. Many
classifiers like KNN, SVM, CNN are used to identify the digit from the
handwritten image. as per the review, CNN is providing better performance than
others. Stages of HDR using CNN classifier is discussed in this paper. MNIST
dataset consist of handwritten numbers from 0-9 and it is a standard dataset used
to find performance of classifiers. HDR consists of three different stages. First is
preprocessing where dataset is converted into binary form and image processing
has been applied on it. Second stage is segmentation where the image is
converted into multiple segments. Third stage is feature extraction where
features of image are identified. Last stage is classification where classifiers like
KNN, SVM, CNN are used. Results of HDR is improved a lot by using CNN classifier
but it can be improved further in terms of complexity, duration of execution and
accuracy of results by making combination of classifiers or using some additional
algorithm with it.
In this paper, they demonstrated residual neural networks. The technology has
evolved over two decades and with the improvement of graphics hardware they
are becoming the industry standard for solving a large number of computer
vision problems, as well as other domains. In particular, the best results reported
in this survey are based on complex neural networks, sometimes combined with
other methods to improve performance. In some cases, CNNs can be combined
with evolutionary algorithms or other optimization techniques to increase the
selection of hyperparameters, allowing better accuracy. In other cases, different
CNN models are combined into one committee, reducing classification error.
Furthermore, we can err that data enhancement techniques improve
performance from surveyed work. These methods are often used with neural
networks or SVMs, and when combined with CNN classifiers, they can top the
entire ranking in the MNIST dataset.
11. Fast Efficient Artificial Neural Network for Handwritten Digit Recognition
By Viragkumar N. Jagtap, Shailendra K. Mishra
In this paper we presented fast efficient artificial neural network for handwritten
digit recognition on GPU to reduce training time with PTM (Parallel Training
Method). We derived back propagation algorithm on GPU based parallelization
should be preferred generally with compared to CPU based program. But still, for
back propagation with small input data and few hidden neurons CPU based
execution is better. But, if the input dataset is larger than GPU based
The result of this study shows that accuracy is improved as the no of blocked are
increased.
Apart from that, a 4x4, 7x7 and 14x14 attribute reduction is performed separately
to compare and find the optimal number of attributes that best represent the
image. The initial hypothesis was answered with concrete data support.
Comparing all the classification models we tested K-Nearest Neighbor is the
preferred choice in terms of its high accuracy and computational efficiency.
However, there is no single classifier that works best on all given problems. A
result shows that probabilistic methods suit better for handwriting recognition. By
varying the training and testing ratios (from 10% to 90%) we found that the larger
training data size improves accuracy, but smaller testing dataset may also favor
better accuracy. Preprocessing such as Attribute reduction (784 reduced to 196)
reduce runtime and increase accuracy (from 93.2% to 95.9%). The proposed
algorithm tries to address both the factors and well in terms of accuracy and time
complexity. The general accuracy of tested K-NN was found to be 96.7% while
96.8% were achieved by Neural Network.
CERTIFICATE
Project Title: <Project Name> Page 21 of 22
TMU-FOE&CS Version 5.0 T004A-Project Report