Speech Enhancement Using Time-Frequency Wavelet Threshold with Auditory Masking
The speech enhancement algorithm based on wavelet threshold is a simple and effective denoising method. In this method, the accuracy of the threshold estimation is crucial to the voice denoising effect. As the diversity of the noise and the time variation of the voice, the usual thresholds can not distinguish noise effectively, and it will damage the voice. In this paper, the threshold estimation in the wavelet threshold is improved. For the different characteristics of voiced speech, unvoiced speech and noise, a multi-threshold dependent on the wavelet packet scale is used. By estimating the signal to noise ratio in different frequency band, the timefrequency adaptive threshold is achieved. Finally, the proposed threshold is adjusted by the auditory masking property. Experimental results show that this method reduces the noise interference, voice distortion is also reduced.
component Speech enhancement Bark wavelet packet transform Noise estimation Wavelet threshold processing Auditory masking
DING Wei WANG Zhong YANG Xiaoyan ZHANG Lei
School of Electrical Engineering and Information, Sichuan University, Chengdu, 610064, China Sichuan Post and Telecommunication College, Chengdu, 610067, China
国际会议
2010 International Conference on Measurement and Control Engineering(2010年IEEE测量与控制工程国际会议 ICMCE2010)
成都
英文
336-340
2010-11-16（万方平台首次上网日期，不代表论文的发表时间）