WSEAS Transactions on Signal Processing
Print ISSN: 1790-5052, E-ISSN: 2224-3488
Volume 15, 2019
An Improved Thresholding Method for Wavelet Denoising of Acoustic Signal
Authors: , , ,
Abstract: To minimize the information loss of acoustic signal and get high SNR in the denoising based on discrete wavelet transform, it is important that the thresholding is suitable for the characteristics of acoustic signal. In this paper, we propose an improved thresholding method to be suitable for the characteristics of acoustic signal. In order to minimize the information loss of acoustic signal in White Gaussian noise, we propose new threshold function to improve the modulus square threshold function. We analyze theoretically a continuity and monotonicity of new threshold function and evaluate the performance of wavelet denoising method based on new thresholding, comparing with Hard, Soft and Modulus square thresholding. Also, we perform the simulation experiment using the various acoustic signals such as mixed acoustic signal of the transient signals, speech signal, shot signal, bird’s song signal and sound signal of gun. The results of theoretical analysis for an improved thresholding show that new threshold function solves the problems of constant error and discontinuity, and minimizes the information loss of acoustic signal. The results of simulation experiment show that SNR of an improved thresholding is highest but RMSE and Entropy are smallest. The theoretical analysis and simulation experiments show that an improved thresholding is more appropriate for acoustic signal denoising based on discrete wavelet transform than previous methods.
Search Articles
Keywords: discrete wavelet transform, non-stationary signal, modulus square threshold function, wavelet thresholding, threshold value, acoustic signal
Pages: 65-76
WSEAS Transactions on Signal Processing, ISSN / E-ISSN: 1790-5052 / 2224-3488, Volume 15, 2019, Art. #9