International Symposium on Chinese Spoken Language Processing (ISCSLP 2002)

Taipei, Taiwan
August 23-24, 2002

Speech Enhancement Using Wavelet Transform with Constrained Thresholds

Ching-Ta Lu, Hsiao-Chuan Wang

Department of Electrical Engineering, National Tsing Hua University, Hsinchu, Taiwan

Although the power spectral subtraction (PSS) in frequency domain and the soft/hard thresholding function in wavelet domain have been successfully applied in speech enhancement, however they increase the probability of negative error and generate musical residual noise. In this paper, we propose a novel scheme that adapts the weight to wavelet coefficients (WCs) for each subband. The weights of all subbands form a weighting function, it can track the speech variation in each subband. Multiplying the weighting function to WCs, the enhanced speech is obtained by inverse wavelet transforming (IWT) the modified WCs. The experiments show that the proposed speech enhancement method can efficiently eliminate the additive noise. Moreover, it is almost free from the musical residual noise.

Full Paper

Bibliographic reference.  Lu, Ching-Ta / Wang, Hsiao-Chuan (2002): "Speech enhancement using wavelet transform with constrained thresholds", In ISCSLP 2002, paper 39.