改进相位谱补偿的语音增强算法  

Speech enhancement using improved phase spectrum compensation

作　　者：王栋[1] 贾海蓉[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030600

出　　处：《西安电子科技大学学报》2017年第3期83-88,共6页Journal of Xidian University

基　　金：国家自然科学基金资助项目(61371193);山西省青年科技研究基金资助项目(2013021016-1);山西省自然科学基金资助项目(2013011016-1);校基金团队资助项目(2014TD028;2014TD029)

摘　　要：针对传统的相位谱补偿方法在低信噪比情况下增强效果不佳的问题,提出了一种改进相位谱补偿的语音增强算法.首先,该算法通过计算每帧语音输入信噪比的相位补偿函数对语音频谱进行补偿;其次,通过新的基于语音存在概率算法估计噪声功率谱密度;最后,把新的相位谱和估计的噪声应用在维纳滤波中.仿真实验表明,文中提出的改进算法能够有效地提高语音系统的去除噪声能力,在低信噪比情况下语音增强效果尤为显著.The typical phase spectrum compensation method has the negative enhancement performance in a low SNR, so the improved phase spectrum compensation method is proposed for this problem. First, the algorithm compensates the speech spectrum through the phase compensation function obtained by calculating the signal to noise ratio of each frame; second, by the new speech presence probability algorithm to estimate the noise power spectral density; finally, we apply the new phase spectrum and the estimated noise in the wiener filter. Simulation results show that the improved algorithm proposed in this paper can effectively improve the ability of voice systems to remove noise especially in a low SNR.

关 键 词：相位谱补偿 语音存在概率 噪声功率谱 维纳滤波 语音增强 

分 类 号：TN912.35[电子电信—通信与信息系统]