基于模糊控制的小波包多阈值语音减噪新算法

A New Algorithm of Wavelet Packet Multi Threshold Speech Denoising Based on Fuzzy Control

在分析小波包传统阈值去噪算法的基础上,提出了一种基于模糊控制的小波包多阈值语音减噪新算法。该算法采用改进的多阈值选取方式来代替传统的阈值选择;应用一种新阈值函数对经小波包变换后的最底层频率系数进行量化处理以确保噪声尽可能地被滤除;模糊控制器可用于对信号中的幅值跳变以及边缘粗糙等问题进行优化与修正。综合以上3种方法即可自适应地进行语音增强处理。经实验结果验证,与传统阈值算法相比,该算法能够最大程度地还原纯语音信息,有效提高了语音去噪的准确度与信噪比。

Based on the analysis of wavelet packet traditional threshold denoising algorithm,a new algorithm of wavelet packet multi-threshold speech denoising based on fuzzy control is proposed. The algorithm adopted an improved multi threshold selection method instead of the traditional threshold selection;applied a new threshold function to quantize the bottom frequency coefficients after wavelet packet transform to ensure that the noise signal can be completely filtered out;fuzzy controller can be used to signal amplitude jump and edge roughness and other issues to optimize and correct. By combining these three methods,the speech enhancement process can be carried out adaptively. The experimental results show that compared with the traditional threshold algorithm,this algorithm restore the pure speech information to the greatest extent,effectively improve the accuracy of speech denoising and the signal-to-noise ratio.