摘要
针对语音增强算法残留"音乐噪声"的问题,分析了基于先验信噪比估计的语音增强算法,并在此基础上提出自适应先验信噪比估计与增益平滑相结合的方法。这种方法先对先验信噪比进行估计,然后对增益函数进行平滑,减小相邻增益函数的随机跳变,弥补了传统先验信噪比估计的不足。最后对含高斯白噪声的语音信号进行处理,仿真结果表明,该算法在抑制"音乐噪声"的效果上得到一定改善,提高了语音增强的性能。
According to the residual "musical noise" phenomenon remaining in the speech enhancement, this paper analyzed the speech enhancement based on the priori, SNR estimation, proposed an algorithm of combining adaptive priori, SNR estimation with smoothing gain, that is, firstly, estimated the A priori SNR, calculate the gain function, then, smoothed the gain function, reduced the gain function of the random neighbor hopping, modified the perceptual of the priori, SNR estimation. Finally, the speech signals with Gaussian white noise were processed by the traditional spectral comparison algorithm and the proposed algorithm. The experimental results show that the proposed algorithm not only effectively suppresses the "musical noise", but also further improves the performance of the noisy speech enhancement.
出处
《计算机应用》
CSCD
北大核心
2012年第A01期29-31,35,共4页
journal of Computer Applications
关键词
语音增强
噪声谱估计
先验信噪比
增益平滑
speech enhancement
noise spectrum estimation
priori SNR
smoothing gain