双通道能量差后滤波语音增强算法统计分析和改进

发布时间：2018-01-21 19:35

本文关键词： 双通道后滤波能量差相干性　出处：《信号处理》2014年07期 　论文类型：期刊论文

【摘要】：双通道能量差后滤波语音增强算法在语音通信系统的噪声抑制技术中有较好的应用前景,然而其理论性能和局限性还未得到充分研究。为此,本文采用统计分析方法研究了双通道能量差后滤波语音增强算法的性能,分析了相干性、平滑因子及噪声估计误差对算法的影响。理论和仿真结果表明,噪声估计误差和平滑因子严重影响该算法的降噪性能。依据此分析结果,本文提出一种基于非平稳噪声估计和功率谱自适应平滑的双通道能量差后滤波算法。测试结果表明,本文提出的算法在不增加语音失真的前提下,能更有效地抑制非平稳噪声,段信噪比提高(SegSNRI)和语音质量感知评估(PESQ)等客观评价指标都表明本文的算法优于其他几种经典的后滤波算法。
[Abstract]:The dual-channel post-filter speech enhancement algorithm has a good application prospect in the speech communication system noise suppression technology, but its theoretical performance and limitations have not been fully studied. In this paper, we use statistical analysis method to study the performance of two-channel post-filter speech enhancement algorithm, and analyze the effects of coherence, smoothing factor and noise estimation error on the algorithm. The theoretical and simulation results show that. The noise estimation error and smoothing factor seriously affect the performance of the algorithm. This paper presents a two-channel energy difference post-filtering algorithm based on non-stationary noise estimation and power spectrum adaptive smoothing. The test results show that the proposed algorithm does not increase the speech distortion. It can restrain non-stationary noise more effectively. The objective evaluation indexes such as segment SNR (SegSNRI) and speech quality perception evaluation (PESQ) show that the proposed algorithm is superior to other classical post-filtering algorithms.
【作者单位】：中国科学院声学研究所通信声学实验室;中国科学院上海高等研究院声学与信息技术研究中心;
【基金】：国家自然科学基金项目(61201403,61302126)
【分类号】：TN912.3
【正文快照】： 1引言语音增强是语音信号处理的重要组成部分,在语音识别、语音编码以及语音通信领域中都有广泛的应用。语音增强是降低噪声干扰,提高通话清晰度和可懂度的有效手段,按通道数目可分为单通道语音增强[1]和多通道语音增强[2],其中多通道语音增强在抑制非平稳噪声和提高语音可懂

【相似文献】