Design of a Low-power, Low-cost ECG & EMG

Sensor for Wearable Biometric and Medical

Application

Abhishek B. Jani (Member, IEEE), Ravi Bagree (Member, IEEE) and Anil K. Roy (Senior Member, IEEE)
Dhirubhai Ambani Institute of Information and Communication Technology, Gandhinagar
Gujarat, India
abhishekbjani@gmail.com; bagree.ravi@ieee.org; anilkroy@ieee.org

Abstract—With the advent of IoT and growing health This paper presents a novel approach towards the design
awareness, the applications of wearable ECG & EMG sensors of an analog frontend for the wearable ECG & EMG
have grown manifold. These applications demand the sensors applications. [2, 4-5]. It is a single board design which uses
to be low-cost, low-power and highly portable. These inexpensive and readily available discrete components. The
requirements put several limitations on the wearable ECG & sensor captures the ECG/EMG signals non-invasively and it
EMG sensors design and development. This paper presents a can send it to the ADC of a microcontroller for post-
new ECG & EMG sensor which had power consumption less processing and analysis. This makes it a multipurpose sensor
than 1.65 mW (I<500 µA @ 3.3 V) and cost less than USD 10. which can be used in several applications including eye
The paper also discusses the implementation details and
blink, muscle movements and for controlling prosthesis
various testing results.
instruments, in a cost-effective manner.
Keywords— Electromyography (EMG), Electrocardiography Section II presents the specifications used for designing
(ECG), Wearable sensor, Wearable device, Sensor based medical the sensor. The details and functionality of various
application, Low power ECG, Low power EMG, Pre-gelled components used in the sensor are presented in Section III.
Electrodes, Instrumentation amplifier. Section IV discusses the design implementation and results
I. INTRODUCTION of some of the experiments.
Development of wearable devices has grown so II. SENSOR SPECIFICATIONS
phenomenally during the last decade that it has got a new The desired specifications of ECG & EMG sensor based
name, Wearable Technology. Besides its fashion quotient, it on the requirements of various wearable applications, viz.,
has serious healthcare applications, particularly for old age heartbeat, eye blink, muscle movements etc. are given in
people and health freaks. The sensors used in such devices Table I.
have several constraints over design and development their
parameters which include wearability, portability, size, TABLE I. SENSOR SPECIFICATIONS
weight, longevity, ergonomics and power consumption [1].
Only after meeting these design parameters such sensors can Parameters Values/Range
be successfully used for continuous health monitoring in Supply Voltage +1.8V to +3.3V
daily life [2]. Current Consumption 300 to 500 µA
Numerous ECG & EMG sensors are already available in Common Mode Rejection Ratio
≥ 100 dB
the market. However, such sensors are either expensive or (CMRR)
not portable. Moreover, majority of them are not based on Gain 100-15000
open-source platform which is desirable to encourage their Low-Frequency Cutoff 0.15 Hz
development and usability for diverse biometric and medical
applications. High-Frequency Cutoff 400 Hz
Weight ≤ 10g
Over the years, various researches have been made
towards the design and application of the ECG & EMG Price ≤ USD 10
sensors. However, most of them do not satisfy criteria such
as, size, operating voltage and power ratings, which are III. SENSOR DESCRIPTION
critical for wearable and battery operated applications. The block diagram of the designed ECG & EMG sensor
Further, these sensors are wire-based which are neither safe is shown in the Fig. 1.
nor reliable for the wearable and portable systems [3].

978-1-5090-1012-7/17/$31.00 ©2017 IEEE

 only. It allowed us to keep the LPF at unity gain, thereby the
linearity and performance of the LPF were maintained
throughout the bandwidth. The gain stage was a simple non-
inverting amplifier followed by a 2nd order inverting multiple
feedback amplifier. We chose 400 Hz as the low pass cutoff
for making the sensor system suitable for multiple
applications. Fig. 2 shows the transfer function of the LPF.

Fig. 1. Sensor Block Diagram

A. Electrodes
The ECG & EMG sensor uses electrodes to acquire the
signals non-invasively. The proposed sensor uses
conventional snap on pre-gelled Ag-AgCl electrodes. This
sticks on the target body and provides good stability to the
system while in motion. This ensures portability of the
wearable device. A pre-gelled electrode also offers high SNR
as it reduces the skin impedance [9].
B. Impedance Matching
An input-stage impedance matching is required to ensure Fig. 2. Transfer Function of the LPF
maximum transfer of power from the electrodes to next
stage. The proposed design uses two basic single ended In last we added a passive anti-aliasing filter just before
passive filters for matching and to resist the flow of any bias the output of the sensor. We used total five op-amps in the
current. It also removes the dc-offset which enables the entire front end, of them two were for the gain amplifier, two
design to achieve high CMRR. This setup also helps in were for LPF and one op-amp for reference generation.
setting a high gain at the instrumentation amplifier stage [1].
IV. SENSOR FABRICATION AND TESTING
C. Instrumentation Amplifier
To implement the readout circuit for the sensor, a two
The sensor uses three op-amp instrumentation amplifiers layer PCB was designed and fabricated. As the sensor is
as it offers various performance benefits compared to other designed for wearable application, the size of the PCB was
configurations. This further improves our system CMRR and an important parameter. The dimension of the PCB was
gain [10]. Further, to remove the common mode signal and 59x15 mm2. It can be reduced further by using professional
EMI noise at 50/60 Hz from the sensor output one can also tools and very small size SMD components. Fig. 3 shows
apply a fix [7] or self-tuned notch filter [2] after the output of the image of the PCB layout. The cost for each sensor is
the amplifier stage. But it also removes critical information calculated approximately to USD 10.
from the signal [11]. Instead, a Right Leg Driver Circuit or
DRL has been used to remove the EMI noise.
D. Driven Right Leg (DRL) Circuit
The human body can act as an active antenna capable of
picking up EMI, especially 50/60Hz noise. To eliminate this
noise the sensor has been designed with Right Leg Driver
Circuit [12]. The DRL is the circuit [13] implemented to
reduce the common mode noise staying on the skin. It also Fig. 3. Sensor Layout
matches the body’s reference potential to that of the circuit.
As shown in the Fig. 1, the DRL stage receives the signal In order to test the sensor and analyze its output, a test-
from the node of Instrumentation Amplifier’s (IN-Amp) gain bench was setup using the Analog Discovery 100MSPS USB
resistors. It can also be obtained from either the gain resistor Oscilloscope & Logic Analyzer of Digilent Inc. make. As
input pins or from the input pins of the amplifier [14]. Thus per the specification mentioned in the Table I, the gain of the
the effective electrode resistance gets reduced by several readout circuit was set to 11400.
orders of magnitude. This allows only a safe amount of Further, a basic skin preparation was done before placing
current flow through the third electrode. the electrodes. It is a standard process to remove dead cells
E. Gain and Low Pass Filter Stage from the skin and to ensure a better connection between the
sensor and the body.
According to the Friis’s formula for noise factor [15],
noise figure increases with each cascaded amplifier stage. Case 1 Sensor is not connected to the body
Therefore, two separate amplifiers, one for improving the
gain and next as a low pass filter (LPF), as shown in Fig. 1 It Fig. 4 shows the output waveform of the sensor when it
resulted in achieving higher gain (> 100) in the initial stage was not placed on the body. The output signal was biased at
the reference voltage of 1.65 V. The measured noise was Sensor Module with EMG & EMG Envelope Output [16] or
1.68 mVp-p. Muscle Signal Sensor EMG Module [17]. This enables the
sensor to work on 3.3 V coin-cell battery for approximately
700 hours. Since the market of wearable biometric and
medical applications is currently in a very nascent stage, we
hope that the ECG & EMG sensor presented in this paper
can open up many new opportunities for the market.
REFERENCES
Fig. 4. Sensor's Output for unconnected leads [1] M. J. Burke and D. T. Gleeson, “A micropower dry-electrode ECG
preamplifier,” Biomedical Engineering, IEEE Transactions on, vol.
Case 2 ECG : Sensor placed in the middle of the chest 47, no. 2, 2000, pp. 155 –162.
[2] M. Nagaraju, J. Silver, F. Zhang, Y. Liao, J. Pandey, T. Morrison, A.
Fig. 5 shows the acquired ECG signal when the sensor Mishra, D. Yeager, and B. Otis, “Circuit techniques for wireless
was placed in the middle of the chest. Its amplitude is 2.30 bioelectrical interfaces,” in 2010 International Symposium on VLSI
Vp-p biased with the reference voltage. The output clearly Design Automation and Test (VLSI-DAT), 2010, pp. 117 –120.
shows all the 6 components of the ECG, i.e., P, Q, R, S, T, [3] A. J. Fridlund and J. T. Cacioppo, “Guidelines for Human
U. Electromyographic Research,” Psychophysiology, vol. 23, no. 5, pp.
567–589, 1986.
[4] R. R. Harrison, H. Fotowat, R. Chan, R. J. Kier, R. Olberg, A.
Leonardo, and F. Gabbiani, “Wireless Neural/EMG Telemetry
Systems for Small Freely Moving Animals,” IEEE Transactions on
Biomedical Circuits and Systems, vol. 5, no. 2, pp. 103 –111, 2011.
[5] R. F. Yazicioglu, P. Merken, R. Puers, and C. Van Hoof, “A 60/spl
mu/W 60 nV/Hz Readout Front End for Portable Biopotential
Acquisition Systems,” in Solid-State Circuits Conference, 2006.
ISSCC 2006. Digest of Technical Papers. IEEE International, 2006,
Fig. 5. Sensor's output from the chest pp. 109 –118.
[6] J. J. Segura-Juarez, D. Cuesta-Frau, L. Samblas-Pena, and M. Aboy,
Case 3 EMG: Sensor placed in the middle of left forearm “A microcontroller-based portable electrocardiograph recorder,”
IEEE Transactions on Biomedical Engineering, vol. 51, no. 9, pp.
Fig. 6 shows the output waveform when the sensor was 1686 –1690, 2004.
placed in the middle of the left forearm. The EMG signals [7] H. I. Cakar, O. Toker, and S. Kara, “A wireless surface
were acquired while moving the wrist up and down in electromyography system design for lumbar disc herniated patients,”
continuous motion. The upper peaks show the event when in 2011 IEEE International Workshop on Medical Measurements and
the wrist was moved in the upward direction, while the lower Applications Proceedings (MeMeA), 2011, pp. 35 –38.
peaks indicate the downward wrist movement. The EMG [8] E. M. Spinelli, R. Pallas-Areny, and M. A. Mayosky, “AC-coupled
front-end for biopotential measurements,” IEEE Transactions on
signals had a peak-to-peak voltage of 2.66 V biased with the Biomedical Engineering, vol. 50, no. 3, pp. 391 –395, 2003.
reference voltage. [9] E. Vavrinsky, M. Daricek, M. Donoval, K. Rendek, F. Horinek, M.
Horniak, and D. Donoval, “Design of EMG wireless sensor system,”
in Applied Electronics (AE), 2011 International Conference on, 2011,
pp. 1 –4.
[10] A. C. MettingVanRijn, A. Peper, and C. A. Grimbergen, “Amplifiers
for bioelectric events: A design with a minimal number of parts,”
Med. Biol. Eng. Comput., vol. 32, no. 3, pp. 305–310, 1994.
[11] C. J. De Luca, L. Donald Gilmore, M. Kuznetsov, and S. H. Roy,
“Filtering the surface EMG signal: Movement artifact and baseline
Fig. 6. Sensor's output for wrist movement when placed on forearm
noise contamination,” Journal of biomechanics, vol. 43, no. 8, p.
1573, 2010.
V. CONCLUSION [12] B. B. Winter and J. G. Webster, “Driven-right-leg circuit design,”
This paper presentd a low-cost, low-power ECG & EMG IEEE Trans. Biomed. Eng., vol. BME-30, no. 1, pp. 62–66, 1983.
sensor which could be used for numerous wearable biometric [13] B. B. Winter and J. G. Webster, “Driven-right-leg circuit design,”
and medical applications. The dimensions of the handmade IEEE Transactions on Biomedical Engineering, vol. BME-30, no. 1,
pp. 62 –66, 1983.
PCB for the sensor readout were 59x15 mm2. The size of the
[14] X. Chen and Z. J. Wang, “Design and Implementation of a Wearable,
PCB can further be reduced to 20x20 mm2 by using small Wireless EEG Recording System,” in (iCBBE) 2011 5th International
size SMD components and professional tools for fabrication. Conference on Bioinformatics and Biomedical Engineering, 2011, pp.
The prototyping cost of each sensor is USD 10. With a gain 1–4.
of 11400, the noise of the sensor was measured to be 1.68 [15] H. T. Friis, “Noise Figures of Radio Receivers,” Proceedings of the
mVp-p. During the experimentation the sensor showed very IRE, vol. 32, no. 7, pp. 419 – 422, 1944.
high sensitivity, capturing and identifying even a small [16] Amazon.com: Tools & Home Improvement [web], Available:
muscle movement quite distinctly. The estimated power https://www.amazon.com/Asiawill%C2%AE-Sensor-Muscle-Signal-
Electromyography/dp/B00LORY62G. Accessed: Jun. 22, 2016.
consumption of the sensor, based on the components
datasheet, is 350 µA. This is significantly lower as compared [17] Amazon.com: Health, Household & Baby Care [web], Available:
https://www.amazon.com/PhysioLab-PSL-iEMG2-Invelope-
to the existing competing devices such as 2 channels EMG Isolation-Available/dp/B0186P3DHM. Accessed: Jun. 22, 2016.