Anda di halaman 1dari 2

ESKILLSTUNA

An Improved LMS Adaptive Filter for Noise Cancellation in ECG Signal


Dandey Venkat Prasad, MD. Anamul Hoque and Abid Aziz Khan
of the patient. Non existence of the equipment to analyse and cancel these more than one at a time is required. Combination of many factors gives rise to research in electrocardiography. They are: 1. Diagnosis of ECG using low-cost high performance technology. 2. Design of portable devices and making the patient less hospital dependent. To develop portable devices and have human computer interaction with microprocessor, there is requirement of less computational algorithm techniques. In this paper improved LMS adaptive filtering technique is applied to cancel the 60 Hz power line interference and baseline wander noise. This is done by selecting better convergence parameter and efficient use of reference signal in discrete time. Simulation is done in MATLAB and the results shows fast convergence and cancelation of noise. III PROBLEM SOLUTION A. Mathematical model: The primary input to the filter is the ECG signal x(k) with additive noise and u(k) is the reference signal which is noise. The error between the initial signal and the product of manipulated version of u(k) assigned as regressor and updated weight w. The error equation is shown below: e(k+1) = x(k) w * regressor. (1)

AbstractNoise cancellation in ECG signal using Adaptive filter are proposed. The mean squared error between the ECG signal which is noisy and a reference input is minimized using adaptive filter. Different filter structures are introduced mainly to cancel the 60 Hz power line interference and baseline wander noise. The adaptive filter works on Least Mean Square (LMS) technique to eliminate the noise that is correlated in some way with the ECG in the primary input. The selection of better convergence parameter and efficient use of discrete time reference signal was done to eliminate the noise in ECG signal. Index Terms Noise cancellation, ECG, Adaptive Filter, Reference signal.

I. INTRODUCTION

HE Electrocardiogram (ECG) recording has a basic role in cardiology. Simple, effective and low cost procedures for the analysis of ECG will have more impact on patient life and also on social costs. Many microprocessor based recorders are invented to do on time signal processing. Due to the limitation of computational power done by microprocessor there is need of digital filters to implement for cancellation of noise. As an example, Lynn [2], [3] and Thakor and Dider [4] have developed integer coefficient and quantized coefficient digital filter respectively for real time execution by microprocessors for ECG analysis. II. STATE OF ART Adaptive filtering technique has become of great importance in many applications related to biomedical field. The fundamental idea about Adaptive filtering has been given in brief by Widrow et al [5]. Many of the ECG processing applications are based on this idea. The idea in [5] is - A reference signal representing power line interference from area not in the vicinity of the ECG recording is used to cancel the interference of power line from the ECG. The literature in this topic introduces an adaptive filtering technique (Least Mean Square) to eliminate the noise from ECG signal which is simple and effective in mathematical computations. III. RESEARCH QUESTION AND PROBLEM STATEMENT Cardiac rhythm, cardiac ischemia and ventricular hypertrophy are the main disturbances in ECG which is due to interference of other signals from other sources in the vicinity

Minimizing the Mean Square Error results in a filter output that is the best least-squares estimate of signal x(k). The LMS algorithm is an iterative technique for minimizing the MSE between the primary and the reference inputs. The LMS algorithm is used to obtain the filter coefficients or weights. The algorithm is written as w(k+1) = w(k) + * regressor. (2) is selected

where w(k) is set of filter weights at time k, randomly to produce convergence at desired rate. Time constant for convergence is1 (4 ).

For stability the parameter is chosen between 1/ and 0. is the largest eigenvalue of the auto correlation matrix of the reference signal.

Manuscript sent October 18, 2008. .

13

ESKILLSTUNA IV. FILTER IMPLEMENTATION IN MATLAB . The main task is cancellation of the diverse noise contributions to ambulatory ECG. Below topics reviews the design and application considered in this paper. A. Baseline Wander and 60 Hz power line noise Reduction The improved LMS algorithm based adaptive filter is simulated in MATLAB. In MATLAB the predefined ECG signal is taken which is given by function ecg (500) which is of 500 Hz frequency. x(k) is the signal which is added ECG signal with noise u(k). Again u(k) is taken as reference signal to find error. Weight of the filter is update as shown in (2). Error is found as shown in (1). is selected as 0.005 randomly changing it between 1/ and 0. The flow chart for the iteration process is shown below. s tart N = 1000 b)

c)

Reference noise signal

Input signal (ECG with Noise)

Display i/p signal

Fig 2: a) ECG signal b) ECG signal with noise and c) ECG signal after filtering Due to better selection of has given deviations for a short time and it reached convergence faster and the original form ECG signal is obtained which noise free . VI. CONCLUSION The achievement of this paper is elimination of base line wander noise using Adaptive Filter using improved LMS algorithm. The important feature of this filter is, without past information the fundamental signal is estimated and is easy to implement on modern microprocessors with numeric abilities. In ambulatory ECG signal many forms of noise occur unpredictably. So, there is a need of multistage filter structure to eliminate the all forms of noise. There is also a need of Artificial Neural Network tools for understanding categories that are able to deal even with high sensitivity to noise and ambiguous patterns. REFERENCES
[1] N. V. Thakor and Yi. Sheng Zhu, Applications of Adaptive Filtering to ECG Analysis: Noise Cancellation and Arrhythmia Detection, lEEE Trans. Biomcd. Eng.. vol. 38, pp. 785 - 794 [2] P. A. Lynn. Recursive digital filters for biological signals, Med. Biol. Eng... vol. 9, pp. 37-43. 1979. [3] -. Transversal resonator digital filters: Fast and flexible on line processors for biological signals. Mrd. Biol. Eng. Cotput., vol. 21. pp. 718-730. 1983. [4] N. V. Thakor and D. Moreau, Design and analysis of quantized coefficient digital filters: Application to biomedical signal processing with microprocessors, Med. Bid. Eng. Cornput., vol. 25. pp. 18-25. 1987. [5] B. Widrow, J. R. Glover, J. M. McCool, et al., Adaptive noise cancelling: principles and applications. Proc. IEEE. vol. 63, pp. 1692-1716, 1975. [6] M. Yelderman, B. Widrow, J. M. Ciotfi, E. Hesler. And J. A. Leddy. ECG enhancement by adaptive cancellation of electrosurgical interference, lEEE Trans. Biomcd. Eng.. vol. BME-30, pp. 392-398,1983.

Manipulate the shifted signal with flipud command

Update weight vector

Error b/n x(k) and noise

End

No

N < 1000

Yes

Fig1: data flow diagram for LMS algorithm

V. RESULTS a)

14

Anda mungkin juga menyukai