基于遗传算法的仿生小波语音增强

Bionic Wavelet Speech Enhancement Based on Genetic Algorithm

分析遗传算法和仿生小波变换的原理和方法,提出一种基于遗传算法的仿生小波语音增强算法。首先将普通小波变换转换为仿生小波变换,得到仿生小波变换系数,接着利用遗传算法的选择、交叉、变异获得仿生小波的优化阈值参数,从而确定最优小波阈值,随后结合最优小波阈值和改进阈值函数去噪,最终将经阈值处理后的仿生小波的系数变换至普通小波域且实行连续小波逆变换,获得增强的语音信号。仿真结果表明,在低信噪比环境下,与传统的最小统计和仿生小波变换算法相比较,经本文提出的算法处理后的增强语音其失真和残余噪声更小,语音质量和可懂度都较高。

By analyzing the principle and method sof genetic algorithm and bionic wavelet transform,a kind of bionic wavelet speech enhancement algorithm based on genetic algorithm is proposed.Firstly,ordinary wavelet transform is converted to bionic wavelet transform and the bionic wavelet transform coefficient is obtained,then the selection,crossover and mutation of genetic algorithm are used for optimizing the bionic wavelet threshold parameter to determine the optimal wavelet threshold value.Secondly,the improved threshold function is combined with the optimal wavelet threshold for denoising.Finally,coefficient of the bionic wavelet after threshold processing is transformed to the ordinary wavelet domain and implements continuous wavelet inverse transformation,enhanced speech signal can be obtained.Comparing with the traditional minimum statistics and bionic wavelet transform algorithm,the simulation results show that in low SNR environment,the enhanced speech processed by proposed algorithm has smaller distortion and residual noise,as well as high voice quality and intelligibility.