含噪语音实时迭代维纳滤波

摘要

针对传统去噪方法在强背景噪声情况下,提取声音信号的能力变弱甚至失效与对不同噪声环境适应性差,提出了达代维纳滤波声音信号特征提取方法.给出了语音噪声频谱与功率谱信噪比迭代更新机制与具体实施方案.实验仿真表明,该算法能有效地去噪滤波,显著地提高语音识别系统性能,且在不同的噪声环境和信噪比条件下具有鲁棒性.该算法计算代价小,简单易实现,适用于嵌入式语音识别系统.%As many traditional de-noising methods fail in the intensive noises environment and be unadaptable in various noisy environments, a method based on iterative Wiener filtering feature extraction is applied for acoustic signals.It frames the acoustic signals at first.Then, the iterative-renewing methods are advanced in noising spectral frequency and Sig-nal-to-Noise ratio (SNR) of spectral power.This method is implemented in detail.The experimental results show that the proposed algorithm can filter noise from voice effectively and improve the performance of automatic speech recognition system significantly.lt is proved to be robust under various noisy environments and SNR conditions.The algorithm is of low computational complexity which is suitable for embedded automatic speech recognition system application.