Enhancement of Non-Air Conducted Speech Based on Wavelet-Packet Adaptive Threshold
Sheng Li,Huijun Xue,Guohua Lu,Yang Zhang,Teng Jiao,Jianqi Wang,Xijing Jing
DOI: https://doi.org/10.11591/telkomnika.v11i1.1879
2013-01-01
Abstract:This study developed a new kind of speech detecting method by using millimeter wave. Because of the advantage of the millimeter wave, this speech detecting method has great potential application and may provide some exciting possibility for wide applications. However, the MMW conduct speech is in less intelligible and poor audibility since it is corrupted by additive combined noise. This paper, therefore, also developed an algorithm of wavelet packet threshold by using hard threshold and soft threshold for removing noise based on the good capability of wavelet packet for analyzing time-frequency signal. Comparing to traditional speech enhancement algorithm, the results from both simulation and listening evaluation suggest that the proposed algorithm takes on a better performance on noise removing while the distortion of MMW radar speech remains acceptable, the enhanced speech also sounds more pleasant to human listeners, resulting in improved results over classical speech enhancement algorithms. DOI: http://dx.doi.org/10.11591/telkomnika.v11i1.1879