REDUCING NOISE FROM SPEECH SIGNAL USING WAVELET PACKET TRANSFORM WITH COIF4 AS MOTHER WAVELET

Abstract

Noise is the main factor affecting the accuracy of the results of signal processing, the use of wavelet transforms to reduce the noise contained in the sound signal, whose noise is intentionally entered using white Gaussian noise at several levels for simulation reasons. Furthermore, noise removal techniques become very important in signal processing before being used in other applications. Coiflet Order 4 is used as a wavelet (mother wavelet) function because it is the best wavelet function for processing voice signals, which is then combined with a soft threshold as a thresholding rule for processed signals. A small Mean Square Error (MSE) value is obtained in the decomposition process at level 4, for each level of WGN given to the signal. From the process carried out, the results obtained for noise 00dB, 10dB and 20dB respectively 93.68%, 91.36%, and 76.27%, all processes are carried out using MatLab 2018a.Keywords: denoising, wavelet, decomposition, noise reduction, threshold