Abstracts Submitted: 459
Number of Users: 527
View Abstracts Submitted
Back to home Page
As Electrocardiogram (ECG) is a non-stationary signal, therefore its Fourier transform (FT) does not give good time or frequency resolution simultaneously. Also, it is well-established by the research communities that FT is not the most viable tool for non-stationary signals. Hence joint time-frequency analysis tools are introduced to get more accurate results in non-stationary signal environment. ECG is an important tool for providing information about functional status of the heart. Power line interference, muscle contraction noise, poor electrode contact, patient movement, and baseline wandering are some of the unwanted noises, which generally gets embedded to ECG during its acquisition and transmission, and should be removed at the prior stage. As the performance of ECG analyzing system primarily depends on the accurate and reliable detection of the QRS complex, in addition to T-wave and P-wave, these noises make it difficult for physicians or medical practitioners to evaluate cardiovascular diseases correctly. So in this context, this paper presents a novel approach for ECG signal denoising and QRS localization based on Fractional S-transform (FrST), which is a modified variant of S-transform (ST) in fractional frequency domain. FrST helps to perform operations in transformed domain where the signal is highly concentrated, so that noise can be easily separated out from ECG signal. The proposed transform has given better time and frequency resolution than the prevailing ST. To validate the performance, the proposed denoising method is tested by artificially corrupting the ECG data by adding additive white Gaussian noise generated in MATLAB. ECG data is taken from standard database of MIT-BIH Arrhythmia which comprises of several abnormal and normal ECG signals. The simulation results show better Root Mean Square Error, Percent Root Mean Square Difference, and improved Signal to Noise Ratio values, which proves the superiority of proposed denoising method than the existing denoising methods. Five records with 5000 samples from MIT-BIH database is chosen to show the comparison between the proposed method and existing ST based method. The former method has better QRS localization, thus outperforming the state of art methods in terms of increased Sensitivity and Positive Predictivity. Also, the proposed method has lesser number of false positives and false negatives.
© Copyright 2017 All Rights Reserved