Heart Sound De-Noising Using Wavelet and Empirical Mode Decomposition Based Thresholding Methods

Heart sound de-noising is considered as an important signal pre-processing step in developing computer assisted heart auscultation model. In this paper, we investigate three white noise reduction methods, namely wavelet transform, wavelet packet transform, and empirical mode decomposition for heart sound de-noising. The de-noised signals are evaluated using signal-to-noise ratio and root mean square error. The results show wavelet transform and empirical mode decomposition methods outperform the wavelet packet transform in heart sound de-noising. The wavelet transform method with 'dmey' wavelet provides a better result for most of the heart sound records. These three de-noising methods are useful to attenuate the white Gaussian noise. It can provide a high quality signal for further signal processing and classifying the heart sound signal.