PERBANDINGAN MOTHER WAVELET DALAM PROSES DENOISING PADA SUARA; THE COMPARISON OF MOTHER WAVELET IN DENOISING PROCESS ON SPEECH

RAHMAT RAMADHAN, Agfianto Eko Putra

2013 | Disertasi | PROGRAM STUDI S2 ILMU KOMPUTER

Wavelet Transform is a method that developed for analyzing the nonstationer signal. Wavelet Transform was used in denoising process on speech to enhance the quality of speech that courrupted by noise. The kinds of involved noises are White Gaussian Noise (WGN), White Uniform Noise (WUN) dan Colored Noise. In this research, the comparison of mother wavelet is performed among Daubechies, Coiflet and Symlet in denoising process on speech. The thresholding method that used in this denoising process is Soft Thresholding. The threshold value is Time Adapted Threshold (TAT) that obtained by estimating the power was used to generate the signal through Teager Energy Operator (TEO). The kinds of tests that used for obtaining the best moher wavelet is Kruskal- Wallis test and followed by Mann-Whitney test. The result shows that Db20, Db30, Db40 and Coif5 mother wavelets are better than others to reduce WGN; Db40, Db20 dan Db30 mother wavelets are better then the other to reduce WUN and to reduce Colored Noise can use some mother wavelets in this research, except Db30 and Db40

Kata Kunci : Transformasi Wavelet; mother wavelet; denoising; sinyal suara; derau; TEO; TAT; Soft Thresholding