基于离散小波变换与小波包分解的语音增强算法

Algorithm for speech enhancement based on discrete wavelet transform and wavelet packet decomposition

提出了一种基于离散小波变换(DW T)和小波包分解(W PD)的语音增强算法。该算法首先将带噪语音进行离散小波变换,并分别对离散逼近信号和离散细节信号采用不同的基小波进行小波包分解,再按照不同的规则选取阈值进行去噪,最后对去噪后的语音信号完成重构。计算机仿真表明,在计算量相当的情况下,该算法优于离散小波变换法去噪和小波包分解法去噪。

An algorithm based on discrete wavelet transform （DWT） and wavelet packet decomposition （WPD） for speech enhancement was presented. It was described as follows： DWT is first used for noisy speech. WPD was then employed for the discrete approximation coefficients and the discrete detail coefficients by different basis wavelets. Various rules were adopted for selecting thresholds to denoise. The denoised signal was finally reconstructed. Computer simulation results indicate that the new algorithm can obtain better performance than those of DWT and WPD under comparative computational complexity.