Printer Friendly

Wearable Electrocardiogram recording using Application-specific Integrated Circuit (ASIC).


Measurement and monitoring of ECG signal a vital diagnostic instrument for countless Heart diseases. Monitoring of ECG and blood flow, measurement of stroke volume and Cardiac Output (CO) plays a vital act in the diagnosis of extensive illnesses such as hypertension and heart failure. The most common methods for computing Cardiac Output are Invasive intra-cardiac catheterization and non-invasive Doppler ultrasound anatomy (Fortin J. et al, 2006). In early days for continuous measurement and monitoring of Heart Rate the patient must be hospitalized. Also the heart rate of a person under rest varies from the heart rate of the same person under his daily activities. Hence necessary of ambulatory monitoring of ECG signal come in existence. The main advantage of ambulatory monitoring is to prevent the deadly cardiac diseases in future. This leads to the invention of wearable ECG monitoring device (Long Yan and Joonsung Bae., 2011). The use of wearable ECG becomes mandatory to record ECG signals in real time.

The main difficulty in computing the heart rate of a patient in motion is the presence of Motion Artifacts along with the ECG signal. The motion artifacts are unwanted signal present along with the ECG signal which reduces the signal quality and leads to defective diagnosis of heart illness for a cardiologist. Many motion artifact reduction techniques (Massagram W. et al., 2010) are available. The block diagram for ECG measurement and processing of ECG signal is shown in Fig 1. ECG is measured using Lead I, II, III through ECG electrode from the patient and is fed as input to an instrumentation amplifier having high CMRR (Ali Hakan Isik and Osman Ozkaraca, 2011) to remove fluctuations in the ECG signal. ADC is provided to convert analog ECG signal into digital output. Since the proposed method mainly focuses on Least Mean Square adaptive filtering algorithm to reduce the motion artifacts with low power, ECG database is collected from PhysioNet and simulated using MATLAB and converted into text file in binary format. The binary file is given as input to the VLSI architecture for LMS algorithm.

Motion Artifact Reduction Using Adaptive Filters:

Artifacts transpire due to movement of electrode that is allocated alongside patient skin. These artifacts encompass of larger amplitude and are comparable to ECG signals. The Artifacts have comparable frequency spectrum to the target bio potential signals. Ambulatory recording systems tolerate from motion artifacts leads to fake alarm and wrong event detection. These motion artifacts can be cancelled from ECG signals by employing adaptive filter for achieving reliable and elevated integrity recording quality under ambulatory conditions. Adaptive filter utilizes optimization algorithms to remove motion artifacts in ECG signal. elevated integrity recording quality under ambulatory conditions. Adaptive filter utilizes optimization algorithms to remove motion artifacts in ECG signal.

Least Mean Square algorithm is one of the linear low power consuming adaptive filtering algorithms that are utilized to remove gesture artifacts present alongside the ECG signal. Fig. 2 shows the method of cancelling motion artifacts from ambulatory recorded ECG signal. The primary input to the filter is the ambulatory recorded ECG signal and the reference input is the original or motion artifact free ECG signal. The input signal x(n) and output signal y(n) of the adaptive filter is expressed in (1), (2) and (3).


Memory based Multiplication:

Finite Impulse Response (FIR) is a common tool used in many signal processing algorithms. Digital signal processing algorithms such as Least Mean Square requires more memory based computations to be performed rapidly and repetitively on set of data. Also the number of multiply-accumulate (MAC) operations of FIR filter increases with the filter order. The implementation of such large order filters is extremely difficult. To reduce the computational complexity, memory based multiplications are replaced by Look up Table based multipliers. To reduce the computational complexity, memory based multiplications are replaced by Look up Table based multipliers. Of the two operands, one is fixed coefficient A and the variable input word X of length L is multiplied by A.

Conventional LUT size increases with the input word size. The product word is stored in location [X.sub.i] for 0 [less than or equal to] [X.sub.i] [less than or equal to] [2.sup.L] - 1. L-bit binary value of [X.sub.i] is used as the address for the LUT to store the product word. The LUT size is decreased by employing assorted optimization methods such as Anti-symmetric Product coding (APC), Odd Multiple Storage (OMS).

LMS algorithm with LUT based multiplier optimization techniques Anti-symmetric Product Coding:

In APC technique (Pramod Kumar Meher, 2010), LUT size is reduced half than the conventional LUT. This technique stores only upper half of conventional LUT words. Consider a 5 bit input word X

X = {[x.sub.4][x.sub.3][x.sub.2][x.sub.1][x.sub.0]} (6)

Where [x.sub.4][x.sub.3][x.sub.2][x.sub.1] [x.sub.0] are the four least significant bits of

X = [X.sub.L.sup.'], for [x.sub.4] = 1 (7)

[[X.sub.L], for [x.sub.4] = 0

Product value = 16A + (signed value) * APC (8)

Where sign value = 1 for [x.sub.4] = 1 sign value = -1 for [x.sub.4] = 0

The product words for input word X of Length L=5 is shown in Table I.

A. Odd Multiple Storage:

In OMS technique (Meher, P.K., 2009), only odd multiples are stored in LUT's and even multiples are attained by easy left shifting operation. The Select lines for shifting will be derived from barrel shifter. LUT size is reduced to half from its original. The APC words stored in LUT are shown in Table II.

B. Combined APC-OMS Technique:

The Combination of APC and OMS technique (Meher, P.K., 2009) will reduce the LUT size further by 50% .Overall size reduction will be around 75 % as compared to conventional LUT. The block diagram for APC-OMS technique shown in Fig. 3 consists of a barrel shifter to store even multiple values.

X Generation Module takes a 5-bit as input. It is utilized to produce anti-symmetric of last 4-bits (X (4 to 0)) when the MSB of X i.e. X (4) = 0 and process the same input when the MSB of X= 1 hence only 16 combinations will be achieved for 5-bit of input. If (x (4) = 0') then [X.sub.comps] = X (4) & 2'scomplement of (X (3 to 0)). Else [X.sub.comps] = X. The address generation unit generates the 4-bit address for the input given by X generation module the 4-bit address is named as d. The reset output will be set, when the input combination X =10000 to make the output of the barrel shifter to 0


[bar.([x.sub.0] + [x.sub.1] + [x.sub.2] + [x.sub.3] + [x.sub.4] )x [x.sub.4]] (9)

[X.sup."] is generated by shifting out leading zeros in [X.sup.']. The 5-bit input word X can be mapped into a 4-bit LUT address (d3d2d1d0), by set of mapping relations as

X [= , for [x.sub.4] = 1

[, for [x.sub.4] = 0 (10)

The address for LUT is calculated as in (11). Values from LUT will be shifted based on select lines ([s.sub.1] [s.sub.0]}

[d.sub.i] = [x.sup.".sub.i+1] for i = 0,1,2 } (11)

[d.sub.3] = [[bar.x].sup.".sub.0 }

The select lines for barrel shifter are estimated as in (12). The control and reset circuit can be designed as

[s.sub.0] = ([x.sub.0] + ([x.sub.1] + [x.sub.2])')' } (12)

[s.sub.1] = ([x.sub.0] + [x.sub.1]) }

Reset = ([x.sub.3] and [x.sub.2]) [x.sub.1]

The Add/Sub Unit either add or subtract barrel shifter output to x4. Mid value 16A is added or subtracted from 16A. Initially 8 odd values are stored in LUT and 2A is stored as a ninth value. This is shown in table III.

5. Experimental Results:

ECG database is loaded and simulated in MATLAB is shown in Fig. 4. LMS algorithm with LUT based multipliers optimized using APC-OMS technique is simulated using Modelsim6.4a and Quartus II tool and the simulation results are shown in Fig. 5. The analog ECG signal with motion artifact and desired signal are given as input in the binary text file format. Motion artifact free ECG signal is viewed as analog signal by using this tool. The comparison results of LMS algorithm with LUT based multipliers and LMS algorithm with LUT based multipliers with optimized APC-OMS technique is shown in Table IV.

Conclusion and Future Work:

To notice and stop the fatal heart illnesses in upcoming there is a demand to compute and monitor ECG of a patient in his daily activities. The main setback of computing and monitoring ECG signal for long term is large requirement of power. Also when the patient is in motion, the observed ECG signal consists of motion artifacts that lead to wrong diagnosis of a cardiologist. The proposed LMS algorithm with LUT based multipliers optimized using APC-OMS technique provides motion artifact free ECG signal. Also the optimized algorithm reduces the size of LUT. Thus area and power consumed by motion artifact reduction algorithm is also reduced. The power of ordinary LMS algorithm and LUT optimized algorithm are compared. The software used for simulation is ModelSim QuartusII. Further in the subsequent step a real time ECG signal is generated from a patient using ECG electrode and processed in Analog Front End to remove fluctuations present along with ECG signal. Then a novel area efficient architecture for LMS algorithm with low power is planned to be implemented in FPGA.


Article history:

Received 1 June 2015

Accepted 28 June 2015

Available online 22 July 2015


Ali Hakan Isik and Osman Ozkaraca, 2011. 'Detection, real time processing and monitoring of ECG signal with a wearable system' IEEE Transactions., pp: 423-427.

Emna Zoghlami Ayari and Reinhard Tielert, 2009. 'A Noise Tolerant Method for ECG Signals Feature Extraction and Noise Reduction' IEEE Transactions., pp: 1-4.

Fortin J. et al., 2006. 'Non-invasive beat-to-beat cardiac output monitoring by an improved method of transthoracic bio-impedance measurement' Comput. Biol. Med., 36: 1185-1203.

Hyejung Kim and Sunyoung Kim, 2014. 'A Configurable and Low-Power Mixed Signal SoC for Portable ECG Monitoring Applications' IEEE Transactions on Biomedical circuits., 8(2): 257-267.

Long Yan and Joonsung Bae., 2011. 'A 3.9mW 25-Electrode Reconfigured Sensor for Wearable Cardiac Monitoring System' IEEE journal of solid-state circuits., 46(1): 353-363.

Massagram W. et al., 2010. 'Digital heart-rate variability parameter monitoring and assessment ASIC' IEEE Transactions on Biomed.Circuits Systems, 4(1): 19-2.

Meher, P.K., 2009. 'New approach to LUT implementation and accumulation for memory-based multiplication' in Proc. IEEE ISCAS pp: 453-456.

Meher, P.K., 2009. 'New look-up-table optimizations for memory-based multiplication' Proc. ISIC pp: 663-666.

Ottenbacher J. et al., 2008. 'Reliable Motion Artifact Detection for ECG Monitoring Systems with Dry Electrodes' IEEE EMBS Conf. pp: 1695-1698.

Pramod Kumar Meher, 2010. 'LUT Optimization for Memory-Based Computation' IEEE Transactions on Circuits and Systems-II: on Express Briefs., 57: 4.

(1) R. Ezhilan and (2) A. Balamurugan M.E

(1) Asst. Professor, VinayakaMission's Kirupananda Variyar Engineering College, Salem, India.

(2) Asst. Professor, Vinayaka Mission's Kirupananda Variyar Engineering College, Salem, India.

Corresponding Author: R. Ezhilan, Asst. Professor, Vinayaka Mission's Kirupananda Variyar Engineering College, Salem, India.


Table I: Apc Words For Different Input Values

                                     f f f f
Input   Product   Input   Product    [x'.sub.1]    APC
(X)     Values    (X)     Values     [x'.sub.0]    WOR

00001   A         11111   31A        1111          15A
00100   4A        11100   28A        1110          13A
00101   5A        11011   27A        1101          12A
00110   6A        11010   26A        1011          10A
00111   7A        11001   25A        1010          8A
01000   8A        11000   24A        1001          6A
01011   11A       10101   21A        1000          4A
01100   12A       10100   20A        0111          2A
01110   14A       10010   18A        0010          1A
10000   16A       10000   16A        0000          0

Table II: APC words for OMS Technique

[X'.sub.3]   VALUE      SHIFTS   INPUT X    APC       [d.sub.3]
[X'.sub.2]                                  WORD      [d.sub.2]
[X'.sub.1]                                            [d.sub.1]
[X'.sub.0]                                            [d.sub.0]

0001         A          0
0010         2XA        1
0100         4XA        2
1000         8XA        3        0001       P0=A      0000
0011         3A         0
0110         2X3A       1
1100         4X3A       2        0011       P1=3A     0001
0101         5A         0
1010         2X5A       1        0101       P2=5A     0010
0111         7A         0
1110         2X7A       1        0111       P3=7A     0011
1001         9A         0        1001       P4=9A     0100
1011         11A        0        1011       P5=11A    0101
1101         13A        0        1101       P6=13A    0110
1111         15A        0        1111       P7=15A    0111

Table III: APC words for combined APC-OMS technique

Address      Stored

0000         1A
0001         3A
0010         5A
0011         7A
0100         9A
0101         11A
0110         13A
0111         15A
1000         2A

Table IV: Comparison Results

Methods                              Power

LMS algorithm with LUT               145.05mW
  based multipliers
LMS algorithm with LUT optimized     128.48mW
  using APC-OMS technique
COPYRIGHT 2015 American-Eurasian Network for Scientific Information
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Ezhilan, R.; M.E., A. Balamurugan
Publication:Advances in Natural and Applied Sciences
Article Type:Report
Date:Jul 1, 2015
Previous Article:FPGA based switching noise reduction technique for multiple input DC to DC converter using sigma delta modulation.
Next Article:User perception of library services in engineering colleges of Tamilnadu, Kerala, Karnataka and Andhra Pradesh: a study.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters