改进小波阈值函数的语音增强算法研究

在小波阈值语音增强算法中，阈值函数是一个重要的部分，其直接决定着语音增强效果的好坏，但现存的阈值函数存在着不连续、计算复杂、不同分解层函数形式固定等问题。为了解决上述问题，本文提出了一种可根据小波分解尺度自适应调整，同时具有调整参数的改进连续阈值函数。该阈值函数在小波域对带噪语音信号的小波系数进行处理，通过遗传算法获取最优解，重构处理后的最优小波系数得到增强的语音信号。本文在仿真与真实环境下进行了实验，改进的阈值函数较传统的阈值函数在信噪比、均方误差以及语音信号主观评价三个方面均得到了提升。实验结果表明，改进小波阈值函数的语音增强算法能有效提升语音信号的可懂度和整体质量。 

Research on Speech Enhancement Algorithm Based on Modified Wavelet Threshold Function

Threshold function plays an important role in speech enhancement using wavelet threshold, which directly determines the effect of speech enhancement. However, the existing threshold functions have various deficiencies, such as incontinuity and computing complex and the fixed functional form in different decomposition levels. In order to resolve the above-mentioned problems, a modified continuous threshold function was proposed, which can adjust automatically with the change of wavelet decomposition scale and have adjusted parameters. Wavelet coefficients of the noisy speech signal have been processed using the modified threshold function in wavelet domain, and the optimal solution was acquired by genetic algorithm. Furthermore, the processed optimal wavelet coefficients were utilized to reconstruct the enhancement speech signal .Experiments have been carried out in simulated and real environments. The modified threshold function had advantage over the traditional one in three aspects of signal-to-noise（SNR）and mean-square error（MSE） and subjective evaluation of speech signal . Experimental results reveal that the speech enhancement algorithm using modified threshold function has effectively improved the intelligibility and quality of de-noised speech signal.