Artifacts Removal from ECG Signal Using a Multistage MNLMS Adaptive Algorithm
International Journal of Signal Processing, Image Processing and Pattern Recognition
Electrocardiogram (ECG) signal is the representation of heart's electrical potential. ECG signal is a small amplitude signal and usually contain numerous types of noises which are classified on the basis of their frequency content. Noises that mostly corrupt the ECG signals are Power line interference, Instrumentation noise, External electromagnetic interference, Baseline drift and noise due to the random movement of the body and respirational movement. The removal of such types of noises is
... pes of noises is very essential for better analysis of ECG signals, which results in a better estimation of the human cardiac system. Many algorithms have been developed for the removal of these artifacts from the ECG signals to extract required information. The primary method is to pass the signal containing the noise through fixed or static filters such as high pass, low pass or band pass filters depending upon the nature of the noise. The static filters have fixed filter coefficients which makes it difficult to remove time varying noise from the signals. To overcome this shortcoming of the fixed filters, different adaptive filtering methods have been developed. Since the ECG signal suffers from several artifacts at a time, which makes a single stage adaptive filter unsuitable for multiple noise signals removal. This paper presents a Multistage Modified Normalized Least Mean Square (MNLMS) algorithm for the removal of multiple artifacts from ECG signals. The results of the proposed algorithm are compared with existing adaptive algorithms including Multistage LMS, Multistage NLMS, Multistage RLS, Multistage SDLMS, on the basis of metrics, including Signal to Noise Ratio (SNR), convergence rate and the computational time, which demonstrate the effectiveness of the proposed algorithm. 19 stage filters i.e., filter 1 and filter 2, was chosen to be 10, linear and nonlinear step size was chosen to be 0.006 and 0.005 respectively. The filter order for second stage filter i.e., filter 3 was chosen to be 200, linear and nonlinear step size for filter 3 was chosen to be 0.004. Figures 3 & 4 shows the original ECG signal and ECG signal corrupted by PLI, Baseline wander and EMG type noise. Simulation results for multistage LMS, NLMS, SDLMS, RLS and MNLMS along with mean square error (MSE) are also shown.