CANCELLATION OF ARTIFACTS FROM CARDIAC SIGNALS USING ADAPTIVE FILTER LMS,NLMS AND CSLMS ALGORITHM

Transcription

1 CANCELLATION OF ARTIFACTS FROM CARDIAC SIGNALS USING ADAPTIVE FILTER LMS,NLMS AND CSLMS ALGORITHM Devendra Gupta 1, Rekha Gupta 2 1,2 Electronics Engineering Department, Madhav Institute of Technology & Science, Gwalior ABSTRACT Electrocardiogram signal in medical science is electric representation of the activity of human s heart. Various cardiac diseases can be recognized with the help of ECG signal. The Electrocardiogram signals are affected by various noises from many sources. The PLI (50/60) is one of the major artifacts in Electrocardiogram signals. So, it is essential to minimize the effect of various noises encountered in ECG during the recording process. This research paper presents the minimization of PLI from ECG signal by using the advanced algorithm which is based on adaptive filter technique known as Constrained Stability Least Mean Square. This algorithm is able to minimize the mean square error and provides better signal to noise ratio with decreasing the convergence time as compared to another techniques. For simulation we used the MATLAB software. Keywords- ECG signal, LMS, NLMS, CSLMS, Adaptive Filter I. INTRODUCTION ECG signal is an electrical signal waveform and graphical representation of electromechanical activity of heart system. In medical science various biomedical signals present in the human body, these biomedical signal helps to check the health condition of the patient. Electrocardiogram signal is one of them. The ECG signal is used to detect or recognize the different types of diseases in the body of patient. The flow of blood and many other activities in the heart system supplies a current flow, which produces an electrical field that can be easily detected, in significantly attenuated form at the body surface, via a different sensors and voltage measuring equipments. When the measured electric field signal, taken with electrodes and shown on oscilloscope screen is known as the electrocardiogram. At the time of recording of ECG signal, various types of noises may corrupt this signal. The electrical waveform representation of ECG signal for single cardiac cycle shown in Figure-1 comprising a P wave, a QRS complex and a T wave. P wave occurs due to sequential activation (depolarization) of the right and left atria.qrs complex occurs due to right and left ventricular depolarization. Normal measurement time for QRS is 60 ms- 100ms. T wave occurs due to ventricular repolarisation. U wave occurs due to repolarisation of the papillary muscles. The common type of noises that corrupts the ECG signal are given below. 1. Power line interference. 2. Baseline line drift. 103 P a g e

2 3. Muscle contraction noise. 4. Electromyography noise. 5. Motion artifacts. 6. Instrumentation Noise. The QRS segment is very important and it is mainly observes to check the patient condition. Remove or minimize these interferences is primary in medical application. So if the noise varies in the amplitude or time duration then the segment wave also changes due to variation in amplitude and time recognizing the true condition of the patient is very difficult task. Therefore the overall concern is to process the ECG signal before observation. The main objective is to separate the ECG signal component from the noise corrupted signal so we can easily understood the actual condition of the Patient. Figure:1 ECG SIGNAL II. ALGORITHMS 2.1 LMS Algorithm LMS algorithm is a stochastic gradient descent method in which the FIR filter change or adapt the weight size based on the error that feedback from output of network. Gradient descent is single degree, single order optimization algorithm. In FIR filter the LMS algorithm is used to optimize filter weights, by updating the filter weights in such a manner to converge the optimum filter weight. Let s(n) is the input signal in vector form consist of time delayed input value -----(1) The weight size in vector form for nth symbol is written as -----(2) The steps for implementing the weight size adaptation are given below Step 1.Calculate value of w (n) at time n Step 2.The equation 3 is used to compute the output written as 104 P a g e

3 ----(3) Step 3.calcuate the difference between desired signal and output signal known as error given by the Equation ----(4) here d(n) is the desired response. It is combination of primary signal s(n) and noise signal n1. Step 4. The next weight size of filter using the designed Equation given below (5) here μ denotes the step-size parameter. Figure: 2 Advance Adaptive Structure Figure 2 shows the advance adaptive structure. In this structure of filter the primary signal s(n) and noise signal n1 added in respective dimension then becomes desired signal d(n). The input signal and noise signal are uncorrelated with each other but the noise signal n1 and n2 are correlated in somewhat manner. The Reference noise signal n2 applied at algorithm based FIR filter. The basic block diagram of the adaptive filter shown in fig.3, in this block diagram the error signal e(n) is feedback and used to update the weight size that will helps in better filtration of s (n). Figure: 3 Block Diagram of LMS Algorithm For the Electrocardiogram signal improvement, the input signal s(n) is corrupted with the noise signal n1 and is applied to the filter based on LMS algorithm whose output is y(n). The input signal and noise n1 are uncorrelated with each other, the mean square error (MSE) is, --(6) For best output of filter the mean square error in the filter output signal should be minimum as possible. At low MSE we will get better estimation of s(n). 105 P a g e

4 2.2 NLMS Algorithm The NLMS algorithm is a enhanced form of the LMS algorithm [11, 12]. Equation 7 shows that how NLMS algorithm updates the coefficients of an adaptive filter. The NLMS algorithm has a time-varying step size μ(n). This step size improves the convergence speed of the adaptive filter. In NLMS algorithm the step size is normalized by power of original input signal s(n). ----(7) 2.3 Proposed Algorithm The CSLMS Algorithm perform efficiently in decreasing mean-squared error (EMSE) and increasing signal to noise ratio. The weight update formula in equation form for conventional CSLMS is written as below ---(8) here, is the difference between input signal and delayed input signal, and is the difference between error sequences. The weight adaptation formula can be made efficient by taking a factor M in denominator and multiply the weight increment by a constant step size μ to enhance the speed of the adaptation. The weight updates equation for CSLMS algorithm written as ----(9) Where, The value of parameter M is equal to norm of input signal s(n). This algorithm provides better result at one half value of norm s(n). III. SIMULATION RESULTS The proposed algorithm has been analyzed using several ECG recordings data collected from MIT-BIH arrhythmia database. In simulations both stationary and nonstationary power line interference (PLI) is considered. The arrhythmia data base consist a signal of 10 second. In MATLAB the signal load in text form and gets result in waveform. In simulation 3600 samples of ECG signal and power line interference (50/60 Hz) in respective dimension mixed with the Electrocardiogram signals to measure the performance of the algorithm in terms mean square error and signal to noise ratio (SNR). We perform the experiment on five data records (101, 102, 103, 104 and 105), But the simulation results for record 101 are shown in this paper. For computing the performance of the proposed adaptive filter algorithm measured the SNR and compared with LMS and NLMS algorithm. Table I illustrate the comparison of SNR for LMS, NLMS and CSLMS algorithms. Table II illustrate the comparison of mean square error (MSE) for LMS, NLMS, and CSLMS algorithm. 106 P a g e

5 3.1 PLI Reduction For power line interference (PLI) cancelation one of five data MIT-BIH record number 101 and 102 are taken. The input to the filter is ECG signal corresponds to the data 101 corrupted with power line interference with voltage 1mv and frequency 60Hz. The ECG signal and Power line interference are uncorrelated with each other. The reference signal is another noise signal n2 that somewhat correlated with noise signal n1, the output of the filter is filtered ECG signal. Power line interference is main source of noise in electrocardiogram signal. This occurs due to variation in power supply of Hospital. Because power supply frequency is 50/60 Hz. Hence we takes PLI in 50/60 Hz range. The mean square error (MSE), signal to noise ratio (SNR) corresponding to step size (u), shown in table I and II. Signal to noise ratio corresponds to step size and record no. Table-1 Step Record LMS NLMS CSLMS Size No. (μ) SNR(db) SNR(db) SNR(db) P a g e

6 Step Size (μ) Mean square error corresponds to step size and record no. Table-2 Record No. LMS NLMS CSLMS MSE MSE MSE (1)Verification for Refinement of Signal by LMS (a) Original ECG signal for record no. 101 After removal of baseline drift Noisy ECG signal Output using LMS 108 P a g e

7 (a) Original ECG signal for record no. 102 After removal of baseline drift Noisy ECG signal Output using LMS (2)Verification for Refinement of Signal by NLMS (a) Original ECG signal for record no. 101 After removal of baseline drift Noisy ECG signal Output using NLMS 109 P a g e

8 (a) Original ECG signal for record no. 102 After removal of baseline drift Noisy ECG signal Output using NLMS (3)Verification for Refinement of Signal by CSLMS (a) Original ECG signal for record no. 101 After removal of baseline drift Noisy ECG signal Output using CSLMS 110 P a g e

9 (a) Original ECG signal for record no. 102 After removal of baseline drift Noisy ECG signal Output using CSLMS IV. CONCLUSIONS In this paper we tried to minimize the noise from ECG signal. The CSLMS algorithm based adaptive filter is used to minimize the noise and distortion from real Electrocardiogram signals that obtained from MIT-BIH data base. The adaptation filter which is based on CSLMS algorithm provides better SNR and mean square error as compare to another algorithms like LMS and NLMS. This is achieved by changing the weight update formula. The CSLMS providing better result as compare to the LMS and NLMS algorithm in terms of SNR, mean square error this is shown in tables I and table II. REFERENCES [1]. Comparative Performance Analysis of LMS and NLMS on ECG Signal by using TMS320C6713 DSK Kit International Journal of Computer Applications ( ) Volume 111 No 12, February [2]. Syed Ateequr Rehman, R.Ranjith Kumar,Comparison of Adaptive Filter Algorithms for ECG Signal Enhancement. International journal of Advanced Research in computer and communication Engineering vol.1,issue2,april [3]. Syed Zahurul Islam, Syed Zahidul Islam, Razali Jidin, Mohd. Alauddin Mohd. Ali, Performance Study of Adaptive Filtering Algorithms for Noise Cancellation of ECG Signal,IEEE2009. [4]. M.Z.U.Rahman, S.R ahamed and D.V.R.K. Reddy, Efficient sign based normalized adaptive filtering technique for cancellation of artifacts in ECG signals: Application to wireless biotelemetry,signal Processing, vol.91,pp ,feb, P a g e

10 [5]. Mohammad Zia Ur Rahman, Rafi AhamedShaik, D V Rama Koti Reddy, Adaptive Noise Removal in the ECG using the Block LMS Algorithm IEEE [6]. M.Z.U.Rahman, S.R ahamed and D.V.R.K. Reddy and Y.Sangeeta, Stationary and Non-Stationary noise removal from Cardiac Signals using a Constrained Stability Least Mean Square Algorithm IEEE 2011 [7]. Y. Der Lin, and Y. Hen Hu. Power-line interference detection and suppression in ECG signal processing. Transactions on BiomedicalEngineering, vol. 55, no. 1, pp , 2008 [8]. Soroor Behbahani. Investigation of adaptive filtering for noise cancellation in ECG signals. IEEE-2007; /07. [9]. Hong Wanl, RongshenFul, Li Shi, The Elimination of 50 Hz Power Line Interference from ECG Using a Variable Step Size LMSAdaptive Filtering Algorithm Life Science Journal, 3(4), [10]. Slim Yacoub and KosaiRaoof, Noise Removal from Surface Respiratory EMG Signal, International Journal of Computer, Information, and Systems Science, and Engineering 2:4, [11]. Stacy Finlay, Carrie Klekta and Ernie Packulak, Adaptive Noise Cancellation for ECG Signal, [12]. Nitish V. Thakor, Yi-Sheng Zhu, Applications of Adaptive Filtering to ECG Analysis: Noise Cancellation and Arrhythmia Detection IEEE Transactions on Biomedical Engineering. 18(8). August [13]. S.C.Chan, Z.G.Zhang, Y.Zhou, and Y.Hu, A New Noise-Constrained Normalized Least Mean Squares Adaptive Filtering Algorithm, IEEE [14]. Ching An Lai, NLMS algorithm with decreasing step size for adaptive IIR filters, Signal Processing,vol 82,pp ,2002. [15]. Desmond B. Keenan, Paul Grossman, Adaptive Filtering of Heart Rate Signals for an Improved Measure of Cardiac Autonomic Control. International Journal of Signal Processing [16]. Rolf Limacher, Removal of power line interference from the ECG signal by an Adaptive digital filter, ETH Zurich, laboratory of Electrical Engineering DesignGloriast, Zurich. [17]. John Leis, Digital Signal Processing-A MATLAB-Based Tutorial Approach, Research Studies Press Ltd [18]. Accessed on [19]. Hakan Johansson, Synthesis and Realization of High-Speed Recursive Digital Filters, University of Linkoping, Sweden, [20]. John G. Proakis, Dimitris G. Manolakis Digital Signal Processing, Principles, Algorithm and Applications. [21]. Edward P. Cunningham, Digital Filtering An Introduction, John Wiley & Sons, Inc [22]. Haykin S, Adaptive Filter Theory, 2nd Edition, United Stated of American: Prentice-Hall, Inc [23]. Philip D. Cha, John I. Molinder, Fundamentals of Signals and Systems. 112 P a g e