Professional Documents
Culture Documents
PDF Analog and Algorithm Assisted Ultra Low Power Biosignal Acquisition Systems Venkata Rajesh Pamula Ebook Full Chapter
PDF Analog and Algorithm Assisted Ultra Low Power Biosignal Acquisition Systems Venkata Rajesh Pamula Ebook Full Chapter
https://textbookfull.com/product/tinyml-machine-learning-with-
tensorflow-lite-on-arduino-and-ultra-low-power-
microcontrollers-1st-edition-pete-warden/
https://textbookfull.com/product/teaching-learning-based-
optimization-algorithm-and-its-engineering-applications-1st-
edition-r-venkata-rao/
https://textbookfull.com/product/efficient-design-of-variation-
resilient-ultra-low-energy-digital-processors-hans-reyserhove/
https://textbookfull.com/product/modern-digital-and-analog-
communication-systems-b-p-lathi/
Ultra high Voltage AC DC Power Transmission 1st Edition
Hao Zhou
https://textbookfull.com/product/ultra-high-voltage-ac-dc-power-
transmission-1st-edition-hao-zhou/
https://textbookfull.com/product/knowledge-management-and-
acquisition-for-intelligent-systems-kenichi-yoshida/
https://textbookfull.com/product/intelligent-algorithms-for-
analysis-and-control-of-dynamical-systems-rajesh-kumar/
https://textbookfull.com/product/charge-sharing-sar-adcs-for-low-
voltage-low-power-applications-1st-edition-taimur-rabuske/
https://textbookfull.com/product/arduino-based-embedded-systems-
interfacing-simulation-and-labview-gui-1st-edition-rajesh-singh/
Analog Circuits and Signal Processing
Analog-and-Algorithm-
Assisted Ultra-low Power
Biosignal Acquisition
Systems
Analog Circuits and Signal Processing
Series Editors:
Mohammed Ismail, Dublin, USA
Mohamad Sawan, Montreal, Canada
The Analog Circuits and Signal Processing book series, formerly known as the
Kluwer International Series in Engineering and Computer Science, is a high level
academic and professional series publishing research on the design and applications
of analog integrated circuits and signal processing circuits and systems. Typically
per year we publish between 5–15 research monographs, professional books,
handbooks, edited volumes and textbooks with worldwide distribution to engineers,
researchers, educators, and libraries.
The book series promotes and expedites the dissemination of new research results
and tutorial views in the analog field. There is an exciting and large volume of
research activity in the field worldwide. Researchers are striving to bridge the gap
between classical analog work and recent advances in very large scale integration
(VLSI) technologies with improved analog capabilities. Analog VLSI has been
recognized as a major technology for future information processing. Analog work
is showing signs of dramatic changes with emphasis on interdisciplinary research
efforts combining device/circuit/technology issues. Consequently, new design con-
cepts, strategies and design tools are being unveiled.
Topics of interest include:
Analog Interface Circuits and Systems;
Data converters;
Active-RC, switched-capacitor and continuous-time integrated filters;
Mixed analog/digital VLSI;
Simulation and modeling, mixed-mode simulation;
Analog nonlinear and computational circuits and signal processing;
Analog Artificial Neural Networks/Artificial Intelligence;
Current-mode Signal Processing;
Computer-Aided Design (CAD) tools;
Analog Design in emerging technologies (Scalable CMOS, BiCMOS, GaAs,
heterojunction and floating gate technologies, etc.);
Analog Design for Test;
Integrated sensors and actuators;
Analog Design Automation/Knowledge-based Systems;
Analog VLSI cell libraries;
Analog product development;
RF Front ends, Wireless communications and Microwave Circuits;
Analog behavioral modeling, Analog HDL.
Analog-and-Algorithm-
Assisted Ultra-low Power
Biosignal Acquisition
Systems
123
Venkata Rajesh Pamula Chris Van Hoof
imec, Leuven, Belgium imec, Leuven, Belgium
Marian Verhelst
KU Leuven ESAT-MICAS
Leuven, Belgium
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This quote has been the foundation stone on which the work presented in this book
is based upon. Building ultra-low power biosignal acquisition systems for long-
term monitoring poses significant challenges due to limited power budgets and
volumes of data generated. Yet, it is well known that most of the biosignals are
readily compressible. This led to raising the question, in light of David Brady’s
statement, if the biosignal sensor and readout interfaces are designed optimally to
acquire only the relevant information. Further investigation of this question resulted
in proposing a new paradigm in biosignal acquisition and processing–analog- and
algorithm-assisted approach. This approach enables reducing the data rate early in
the signal processing chain, thereby resulting in significant power savings in the
overall system.
This book, apart from introducing the assisted signal processing architectures for
biosignal acquisition and processing, advances the state of the art in the following
areas:
• Building upon the prior work on adaptive sampling for electrocardiogram, the
work presented in this book describes a nano-power adaptive sampling controller
that achieves 8× reduction in the data rate in analog domain.
• A novel, low-complexity, overlapped window reconstruction algorithm for
improved reconstruction accuracy in compressive sampling framework.
• A feature extraction algorithm and an application specific integrated circuit
(ASIC) implementation thereof that enables accurate heart rate estimation
directly from compressive sampled photoplethysmogram signals. This approach
resulted in 30× power savings in light emitting diode (LED) driver and com-
v
vi Preface
vii
Acknowledgments
My time at imec and KU Leuven was made most enjoyable by people around me.
First and foremost, I would like to sincerely thank my promotor, Professor Chris
Van Hoof. His approach of being available to the student when needed, without
being overbearing, has helped me develop as an individual researcher. His advice
and support have been extremely important to me at all phases of my stay at imec
and more so when I went on leave of absence.
Next, I would like to thank my co-promotor, Professor Marian Verhelst, for she
has been immensely helpful with both technical and nontechnical advice. Professor
Marian closely followed my work since its inception, and without her continued
support, this work would not have seen the light of the day. Without a doubt, the
entire credit, from inception of this book to its current form, goes to her. Professor
Marian also encouraged to actively collaborate with her group, and this has given
me an opportunity to expand my research horizon.
My sincere thanks are also due to Dr. Refet Firat Yazicioglu for agreeing to
serve on my dissertation committee. It was Firat who originally interviewed me and
brought me to imec. He has been supportive during the entire course of the work
presented in this book, and I am highly indebted to him for his contributions to this
work.
Next, I would like to thank my colleagues at imec and KU Leuven. Rachit Mohan
has been a good friend at work. Not only we shared our office but also had engaging
discussions both on technical and nontechnical aspects. Dr. Hyejung Kim has been
a great mentor during my initial days at imec and at later stages Dr. Yan Long.
Dr. Long, in particular, was extremely helpful in developing the PPG SoC, and his
critical feedback has helped me improve on several fronts. Dr. Nick Van Helleputte
has been extremely supportive toward the final stages of this work, and I would like
to extend my thanks to the rest of the colleagues from BAN/MEDIC group. Special
thanks are due to Marion Hegemann, for she has taken care of all the administrative
aspects during my stay at imec. From KU Leuven, Komail Badami has been a
good friend with whom I had several discussions both on and outside work and
collaborated on certain projects. Juan Carlos Pena and Steven Lauwereins also from
KU Leuven have been good friends and partners in several technical discussions.
ix
x Acknowledgments
The administrative department and the international admissions and mobility office
of KU Leuven and the personal department of imec have been extremely helpful
in making relocation to Belgium a smooth process. Heidi Lowet, Kristel Paulissen,
and Bob Geivers have been particularly helpful from the international admissions
and mobility office, while Marianne Van Rompay from the personal department of
imec took care of the necessary administrative processing at imec.
My stay in Leuven has been made memorable, thanks to my friends, some
of whom I got to know quite well during my stay. Rathaiah Pureti and his wife
Sulochana, Ravi Chandra Chintala and his wife Nirosha, Prashanth Agarwal and his
wife Padmaja, and Ajay Kumar Kambham and his wife Sushma Priya have been
extremely friendly and supportive during my stay in Leuven. Other members from
Indian community, Chaitanya Cherukuri and Swapna, Satish Achanta and Swapna,
and Prasanna Chaduvula and Sridevi, have hosted me on several occasions. At imec,
I have had great time interacting with Bharani Chava, Yoga Saripalli, Arul Mahesh
Jagadeesa Das, Aftab Nazir, and Raf Appeltans. Raf and myself worked together
as teaching assistants at KU Leuven for three straight years, and it was a pleasure
working with him. I am extremely thankful to Frieda Hermans, who has been my
landlord during my stay in Leuven, who always made sure that I had a comfortable
stay.
Outside imec and KU Leuven, I would like to thank Professor Shanthi Pavan
and Professor Nagendra Krishnapura from the IIT Madras, India. Their lectures
have been a trove of knowledge and helped me develop deeper insights into circuit
analysis and design. Thanks are also due to Professor Yannis Tsividis of Columbia
University, New York City, USA, who gave valuable insights and directions during
my visit. I would also like to acknowledge the contributions of Jose Manuel
Sarmiento, who was a visiting student at imec from North Carolina State University,
Raleigh, USA, to the PPG SoC. I would also like to extend my thanks to Professor
Visvesh Sathe and Professor Chris Rudell of the University of Washington, Seattle,
USA, for their support and inculcating deeper insight into circuits.
Last but not least, I want to thank my dearest love, my wife, Kamala Purnima
Pisipati. I could not have asked for a better companion, for she has been there for
me always, encouraging and supporting me in all the endeavors that I undertook.
Thanks are also due to my parents, in-laws, sister, and her family for their continued
support and encouragement.
The work presented in this book was made possible by the funding provided
by the European Commission FP7 FET PHIDIAS project (Grant agreement no.
318013). Special thanks are also due to Tobias Vanderhenst of Europractice for
coordinating the tape-out efforts with TSMC.
Contents
xi
xii Contents
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Abbreviations and Acronyms
xiii
xiv Abbreviations and Acronyms
xvii
xviii List of Figures
Fig. 3.10 Average SNDR for traditional approach and the proposed
overlapped window approach over 25 runs across 20 records . . . . . 48
Fig. 3.11 Analog CS implementation through RMPI . . . . . . . . . . . . . . . . . . . . . . . . . 49
Fig. 3.12 Digital CS implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Fig. 3.13 CS implementation through NUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Fig. 3.14 Analog CS for EEG acquisition, exploiting spatial
correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Fig. 4.1 Various possible CS based acquisition systems in context of
BAN. (a) Signal acquisition is performed at the sensor node
while reconstruction and feature extraction is performed at
the base station. (b) Both CS encoding and decoding are
performed on the signal node followed by feature extraction.
(c) Feature extraction is performed on the sensor node
directly from the CS data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Fig. 4.2 Estimation of HR and HRV from time domain PPG signal . . . . . . . . 59
Fig. 4.3 Estimation of average HR from the frequency spectrum of
PPG signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Fig. 4.4 Average error in the inner product of randomly sampled
PPG signal normalized to the corresponding Nyquist rate
inner products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Fig. 4.5 Concept of frequency spectrum estimation from randomly
sampled signal using least-squares spectral fitting . . . . . . . . . . . . . . . . . 61
Fig. 4.6 (a) Four time domain segments of PPG signal each 4 s
long. (b) Estimated spectrum from 10× randomly sampled
signals. α indicates the significance level of the peak in
spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Fig. 4.7 HRV using proposed approach and conventional approach.
HRV computed using proposed technique is highly
correlated to the one computed using conventional time
domain approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Fig. 4.8 FFT of time domain PPG signals in Fig. 4.6a after
downsampling the signal by (a) 10× (b) 30×. DC
component is removed for clear representation . . . . . . . . . . . . . . . . . . . . . 65
Fig. 4.9 Alias creation by downsampling the signal and the absence
of the same when random sub-sampling is employed . . . . . . . . . . . . . . 66
Fig. 4.10 (a) Time domain PPG signal randomly sampled to obtain
30× compression. Sampling instants are highlighted. (b)
Estimated spectrum of the randomly sub-sampled signal . . . . . . . . . . 66
Fig. 5.1 Conventional PPG acquisition system employing uniform
LED stimulation and sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Fig. 5.2 CS based PPG acquisition system employing random LED
stimulation and sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Fig. 5.3 Partial measurement matrix structure for CS PPG acquisition . . . . . 72
xx List of Figures
Fig. 5.4 LED and sampling pulse structure for conventional (uniform
sampling) and CS PPG acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Fig. 5.5 The architecture of a single channel CS PPG acquisition
ASIC which embeds a DBE for feature extraction . . . . . . . . . . . . . . . . . 74
Fig. 5.6 Ideal PD and TIA interface. The TIA is formed by
employing resistive feedback around an OTA . . . . . . . . . . . . . . . . . . . . . . 75
Fig. 5.7 (a) Ideal PD and (b) a practical PD with shunt resistance
(Rp ) and reverse bias junction capacitance (Cp ) . . . . . . . . . . . . . . . . . . . 75
Fig. 5.8 (a) Equivalent circuit for loop gain computation of TIA. (b)
Equivalent circuit for return ratio computation . . . . . . . . . . . . . . . . . . . . . 76
Fig. 5.9 Magnitude response of AOL and the reciprocal of feedback
fraction β1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Fig. 5.10 Lead-lag compensated TIA. Stability margin is enhanced by
inserting a feedback capacitor (Cf ) into the network, which
introduces a left half plane (LHP) zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Fig. 5.11 (a) Equivalent circuit for loop gain computation of the
lead-lag compensated TIA. (b) Equivalent circuit for return
ratio computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Fig. 5.12 Measured reverse bias junction capacitance of the
photodiode used in the current work as a function of reverse
bias voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Fig. 5.13 TIA interfaced with a current DAC (IDAC) at the input.
IDAC enables active rejection of Idc , relaxing the channel
DR requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Fig. 5.14 Schematic of the 5-bit current DAC (IDAC) used to subtract
static component of photocurrent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Fig. 5.15 Two stage Miller compensated OTA with resistive source
degenerated current source in the first stage. This OTA core
is used to realize the TIA and SI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Fig. 5.16 Switched integrator (SI) implementation in the current
work. The output of the TIA is further amplified and filtered
through SI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Fig. 5.17 Switched integrator (SI) as noise limiting filter in the PPG
readout channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Fig. 5.18 Schematic of the 12-bit SAR ADC with a unit capacitor of
800 fF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Fig. 5.19 The simplified architecture of the digital back end (DBE) . . . . . . . . . 85
Fig. 5.20 Key timing signals for AFE and LED driver control . . . . . . . . . . . . . . . 86
Fig. 5.21 Eight-way multiply accumulate (MAC) unit for accelerating
PSD estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Fig. 5.22 Chip micrograph of the CS PPG ASIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Fig. 5.23 (Top) Measured output of the channel for a DC current
excitation of LED in uniform sampling mode. (Bottom)
Zoomed in view of response during one sampling instant . . . . . . . . . 89
List of Figures xxi
Fig. 5.24 Signal acquisition with CRs 8× and 30× when LED is
stimulated with a sinusoidal current at 1.2 Hz . . . . . . . . . . . . . . . . . . . . . . 89
Fig. 5.25 Channel recovery from saturation when IDAC is enabled . . . . . . . . . 90
Fig. 5.26 In vivo acquired PPG signal through the ASIC under
uniform sampling mode and with a CR of 10× . . . . . . . . . . . . . . . . . . . . 90
Fig. 5.27 Measured frequency corresponding to the peak in the PSD
(fpk ) from the ASIC with LED modulated with a sinusoidal
current whose frequency is swept from 0.5–3.4 Hz . . . . . . . . . . . . . . . . 91
Fig. 5.28 Measured power consumption breakdown of the ASIC and
the off-chip LED driver for different CRs . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Fig. 5.29 In vivo acquired PPG signals under different SNR
conditions. The corresponding values of acquired
photocurrent and LED driver current are indicated in
Table 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Fig. 6.1 Concept of motion artifact reduction using spectral
subtraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Fig. 6.2 (a) PPG signal acquired from a subject under normal
office working conditions. (b) PSD of the PPG signal
estimated using LSP after 10× random sub-sampling.
(c) Accelerometer signal acquired simultaneous to PPG
acquisition. (d) PSD of the accelerometer signal estimated
using LSP after 10× random sub-sampling. (e) PSD of the
PPG signal post-spectral subtraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Fig. 6.3 Determination of PAT and HR for cuffless BP estimation . . . . . . . . . 99
Fig. 6.4 ECG assisted PPG acquisition for cuffless BP estimation . . . . . . . . . 100
Fig. 6.5 A 10 s simultaneous ECG and PPG recording obtained
through the COTS platform (average values are equalized
for better representation). Both the signals are sampled at
128 Hz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
List of Tables
xxiii
Chapter 1
Challenges and Opportunities
in Wearable Biomedical Interfaces
1.1 Introduction
The changing life styles of a large section of population has increased the vulnera-
bility to cardiovascular disease (CVD) and mental disorders. This continued trend,
coupled with rapidly aging population, exerts tremendous pressure on the existing
healthcare and financial system. For instance, CVD is the leading cause of deaths
in the Europe, accounting for 47% of total deaths in 2012, costing the economy
e196 billion a year [1]. Similar statistics were reported from the USA, where in
2009, CVD accounted for nearly 32% of all the deaths in the USA, leading to an
estimated cost of $312.6 billion [2]. These increased costs, along with an increasing
population of health-conscious subjects, have resulted in a paradigm shift towards
preventive healthcare, the aim of which is to shift healthcare from the hospital to the
home. The proponents of this paradigm argue that this approach helps in avoiding
people getting sick in the first place or detect the onset of illness at an early stage
in disease cycle, thereby leading to increased quality of life and reduced healthcare
cost [3].
Wearable and implantable medical devices, that are capable of continuously
monitoring the vital parameters, are seen as the engines driving the preventive
healthcare paradigm. Specifically, wearable medical devices can be used for long-
term continuous health and lifestyle monitoring in an unobtrusive manner. These
wearable devices, possessing sensing, signal processing, and communication capa-
bilities, form a body area network (BAN), with each device being a wireless sensor
node (WSN), leading to connected health. To enable widespread usage of such
devices, it is important to maximize their functionality while requiring minimum
intervention from the users, with extended life time, and minimize size and cost.
While a small form factor wearable healthcare device increases the esthetics,
wearability, and comfort, it severely restricts the battery size, putting severe
constraints on the power budget of such devices, thereby necessitating the need
for the design of ultra-low-power circuits and systems. Moreover, the requirement
for long-term monitoring leads to large amounts of data that need to be processed
and transmitted, which results in significant energy consumption. Early reduction
of data rate in the signal processing chain can lead to significant power savings in
the subsequent stages, thereby leading to reduced overall power consumption of the
system [4].
This work deals with architectural and circuit approaches to enable early data
rate reduction for two different biomedical signal acquisition modalities. The first
design deals with an ultra-low-power adaptive sampling controller (ASC) for elec-
trocardiogram (ECG) acquisition, while the second design deals with a low-power
photoplethysmogram (PPG) acquisition system with embedded feature extraction
and reduced stimulation power. The rest of the chapter introduces background
information necessary to put the rest of the chapters in context. Section 1.2 provides
a brief introduction to the low-power biomedical interfaces, to motivate the need
for early data rate reduction, while Sect. 1.4 presents architectural and algorithmic
approaches to reduce the power consumption of the biosignal acquisition system.
Section 1.3 provides background of the signals dealt in this work, namely ECG
and PPG signals. Finally, Sect. 1.5 provides an overview and summarizes the
organization of this work.
Fig. 1.1 Block diagram of a typical biosignal acquisition system and its power breakdown
1.3 Introduction and Overview of Electrocardiogram (ECG) and. . . 3
by the wireless link, which accounts for nearly 84% of the system power [5]. On the
other hand, AFE and the ADC, combined, consume only 0.3% of the system power.
While the power consumption of the memory and DSP put together stands at 1.6%
of the system power, it is still 5× higher than the AFE and ADC power.
Recent advances have resulted in ultra-low-power consumption for the individual
building blocks on a biosignal sensor node. On the AFE front, Harpe et al.
[6] demonstrated an instrumentation amplifier (IA) for ECG acquisition, which
consumes only 1 nW of power and achieves a noise efficiency factor (NEF),1 defined
in [7], of 2.1. The modern ADCs for biomedical applications, on the other hand,
consume power as little as 1.1 nW [6]. The DSP, including static random access
memory (SRAM), is reported to have achieved a performance of 0.3 pJ/bit for
biomedical signal processing [8]. With technology scaling, favoring the reduced
dynamic power consumption in digital circuits, the performance of the DSP is
predicted to improve, favoring increased processing in digital domain. Advances in
radio design resulted in development of radio transceivers that consume energy per
bit as low as 0.33 nJ/bit [9]. Despite such advances, the radio energy consumption
is still three orders of magnitude higher than the DSP, which is higher than the AFE
and ADC energy consumption combined.
Since the energy consumption of the radio scales with the number of bits
transmitted, an effective way of reducing radio power consumption is by performing
data compression, prior to transmission, thereby reducing the number of bits to
be transmitted. This data compression is traditionally performed in the digital
domain [10, 11]. This approach, therefore, relaxes the energy consumption of the
radio, while increasing the DSP power consumption. While technology scaling
seems to favor processing in digital domain due to reduced dynamic power
consumption, increased leakage power, particularly in systems with large memories,
proves to be the bottleneck [12]. Therefore, it is important to explore the possibility
of reducing the data rate early in the signal processing chain, through analog signal
processing techniques, to reduce the number of bits to be stored and processed by
the DSP, thereby reducing both memory footprint and the energy consumption.
As mentioned earlier, this work deals with the acquisition and processing of ECG
and PPG signals in the context of wearable devices for continuous monitoring of
cardiovascular state. Therefore, it is imperative that the aspects of ECG and PPG
that are necessary to provide sufficient background to the reader are introduced.
This introduction is intended to provide details that are relevant for circuit design
1 Noise efficiency factor (NEF) is a quantitative metric that captures the current consumption–noise
trade-off of IAs.
4 1 Challenges and Opportunities in Wearable Biomedical Interfaces
2 Applications that involve stimulation, as is the case with pacemakers, require high BW to capture
pacing pulses.
1.3 Introduction and Overview of Electrocardiogram (ECG) and. . . 5
enable cuffless blood pressure (BP) monitoring [16, 18]. Since the focus of the
current work is in determination of cardiovascular state, determination of HR and
HRV is of paramount importance, while the aspects of cuffless BP monitoring are
described briefly in Chap. 6.
For applications involving HR and HRV, the AC component of the PPG signal,
specifically, the frequency of the AC component contains the relevant information.
As with the case of ECG signals, HR from PPG signals is also computed by
measuring the temporal difference between successive peaks in the PPG signal.
Alternatively, frequency domain techniques can be employed to estimate the
dominant frequency component in the signal, from which HR can be determined.
The peak-to-peak value of the AC component of the photocurrent depends on several
factors including the intensity of the LED light, quantum efficiency of the PD, the
skin tone of the subject, mode of PPG acquisition, and the location at which the
acquisition is being performed [19]. In terms of relative values, the AC component
is typically 1–4% of the DC component in transmission PPG acquisition mode,
while in reflectance mode the AC component is reported to be 10,000 times lower
than the DC component [20].
Although PPG based HR measurement is a well-understood concept, its usage in
continuous monitoring is hindered due to very high power consumption compared
to ECG based methods. Unlike ECG, PPG acquisition requires active stimulation
through LEDs and the power consumption of a typical PPG acquisition system
ranges from few mWs to tens of mWs, dominated by the LED driver power
consumption [17]. Also, the relatively small AC component in comparison to the
DC component of the photocurrent necessitates the need for a low-noise and large
DR readout. Moreover, PPG acquisition is highly susceptible to motion artifacts.
Compressive sampling (CS) based PPG acquisition, with embedded feature
extraction, that enables up to 30 times reduction in the LED driver power consump-
tion, without compromising on the accuracy of estimated HR, is explored in the
current work to mitigate the high power consumption in PPG acquisition systems,
and the details are presented in Chaps. 4 and 5, respectively.
Fig. 1.5 (a) Conventional biosignal acquisition and processing paradigm. (b) Assisted processing
architecture
sub-blocks assist each other, relaxing the specifications and leading to lower-power
consumption, is often referred to as assisted processing architectures [21].
Figure 1.5 shows conventional and assisted processing architecture for biosignal
acquisition and processing (adapted from [21]). In a conventional system, analog
signal processing is restricted to signal conditioning, resulting in high-BW require-
ments for the ADC. A high-speed DSP then processes the output of the ADC, to
extract the relevant information from the same, which is often lower in BW, with
relaxed precision requirements. In an assisted processing paradigm, a programmable
AFE performs optimum amount of analog pre-processing before digitizing and
further processing the data, thereby relaxing the ADC and DSP speed and precision
requirements. The DSP assists the programmable AFE by rejecting the aggressors
such as motion artifacts [22] and/or power line interference (PLI) [23]. This in turn
relaxes the DR requirements of the AFE, lowering its power consumption. These
architectures, where the DSP assists the AFE to relax its specifications, are often
referred to as digitally assisted analog signal processing architectures. However, it
is must be noted that the digitally assisted analog signal processing paradigm shifts
the computation burden onto the DSP, whose power consumption, as discussed in
Sect. 1.2, depends on the number of bits to be processed.
Analog assisted digital signal processing is a paradigm in which analog pre-
processing is performed to reduce the amount of data to be processed by the DSP.
Analog pre-processing, depending on the application, ranges from simple filtering
and rejection of out-of-band aggressors, extracting the signals in the frequency
8 1 Challenges and Opportunities in Wearable Biomedical Interfaces
This book presents analog assisted signal processing architectures for biomedical
signal acquisition and processing for two physiological signals. The first is an
adaptive sampling approach applied to ECG acquisition, which is described in
Chap. 2. The second is a compressive sampling (CS) PPG acquisition system with
embedded feature extraction, the details of which are described in Chaps. 4 and 5.
The objective of both systems is reliable HR and its variability estimation in a
low-power manner. This objective is achieved through a combination of algorithm
exploration and circuit optimization. The organization of the rest of the book is as
follows:
1. Adaptive Sampling for Ultra-Low-Power Electrocardiogram (ECG) Read-
outs: Chapter 2 presents an adaptive sampling approach for early data rate
reduction in ECG acquisition systems. Adaptive sampling approach is demon-
strated to offer better trade-off between the reconstructed signal quality and the
CR compared to other data rate reduction techniques applied to ECG. An ultra-
low-power adaptive sampling controller (ASC), based on first derivative based
Another random document with
no related content on Scribd:
Colombo, de Cerlone, es México la primera tierra descubierta por el
gran navegante, y, lo que es más, bautizada por éste con el nombre
de América, reconociendo que Américo Vespucio había tenido
noticia antes que él de tal tierra. Semejantes desatinos sólo pueden
compararse con los que contiene el drama Columbus del alemán
Dedekind, publicado en Leipzig este año, entre los cuales sobresale,
sin duda, en primer término la originalísima especie de que la
amistad de Colón con los Pinzones tuvo origen en la Universidad de
Pavía, siendo condiscípulos en sus aulas el marino genovés y los
célebres pilotos de Palos.
En comparación con tales dislates, insignificantes resultan los
errores históricos de los dramaturgos españoles, comenzando por
frey Lope Félix de Vega Carpio.
El Nuevo Mundo descubierto por Christoual Colon es anterior á
1604, fecha en que había sido ya escrita esta comedia, como lo
prueba el hecho de venir incluída en la lista publicada en El
Peregrino, dada á luz en aquel año. Diez más tarde, en 1614, fué
impresa en la Cuarta parte de comedias de Lope. Como hasta
ahora, que sepamos, no ha merecido estudio especial, bueno será
dar alguna noticia de ella con ocasión del presente Centenario.
Al descubrimiento de América se refieren propiamente, en esta
obra, la mayor parte del acto primero, las primeras escenas del
segundo y las últimas del tercero. Las demás tienen por asunto, ya
la conquista de Granada, ya los compañeros que dejó Colón en las
primeras tierras descubiertas entre los cuales incluye Lope á D.
Bartolomé Colón y al P. Buil, que no fueron en el primer viaje, como
es notorio. Las escenas de esta parte abarcan más de la mitad del
acto segundo y casi todo el tercero. La celebración de la primera
Misa, y los amores y amoríos de indias é indios, indias y españoles,
groseramente presentados, les sirven de argumento. Nada de esto
se refiere, pues, directamente á Colón ni á la empresa descubridora.
Tratemos de ésta únicamente. Y desde luego diremos que, desde
el principio hasta el fin, las escenas de nuestra comedia se ajustan
esencialmente á la historia del descubrimiento contenida en las
crónicas españolas que corrían por entonces. El poeta tomó de unas
y otras lo que tuvo á bien escoger, permitiéndose luego algunos
ensanches y alteraciones. Con razón podríamos, pues, calificar su
obra de crónica dialogada, lo mismo en lo que toca á los
preliminares del descubrimiento, que en lo relativo al primer viaje,
que en lo tocante al regreso y presentación de Colón á los Reyes en
Barcelona.
Ni las mocedades de Colón, ni sus amores con D.ª Beatriz
Enríquez, explotados después por otros dramaturgos, así como
tampoco la supuesta intervención de la Universidad de Salamanca
en las negociaciones colombinas, y menos aún las decantadas
burlas de sus maestros de los proyectos del gran navegante, figuran
en modo alguno en nuestra comedia. Lope nos presenta solamente
al descubridor en los tres momentos capitales de su empresa:
buscando favorecedores hasta que los encuentra en los Reyes
Católicos; llevando á cabo su inmortal viaje; regresando vencedor á
España á noticiar su triunfo.
De estos tres momentos, el primero es el tratado con mayor
extensión en nuestra comedia. Colón, viviendo en la isla de la
Madera, hospeda en su casa á un piloto, el cual, al morir, en premio
de la hospitalidad recibida, le entrega sus papeles, y con ellos el
secreto de la existencia de un Nuevo Mundo. Porque es de saber
que aquel piloto, navegando por el mar Océano en una carabela
empujada por contrarios vientos y arrastrada por las corrientes,
había arribado á América. Refiérelo á Colón, añadiendo que
No sé cómo te he escuchado,
Colón, sin auer reído,
Hasta el fin, lo que has hablado;
El hombre más loco has sido
Que el cielo ha visto y criado
Un muerto con frenesí
Te pudo mover ansí
Con dos borrados papeles:
Si de engañar vivir sueles,
¡Cómo te atreves á mí!
Á Castilla y á León
Nuevo mundo dió Colón,