基于改进谐波恢复算法的语音增强方法

Speech Enhancement Method Based on Improved Harmonic Retrieval Algorithm

当信噪比较低时,语音信号的高次谐波部分会完全淹没在噪音中。针对该情况,提出一种基于改进谐波恢复算法的语音增强方法。对经过MMSE-LSA算法语音增强处理后的时域输出语音信号进行非线性处理,得到准周期冲激信号,并将其与原增强信号相乘,突出语音的谐波分量。实验结果表明,改进算法较好地解决了低信噪比时谐波失真的问题,相比传统谐波恢复算法能更好地改善语音高次谐波的质量。

The higher harmonic part of the speech signal is completely overwhelmed by the noise when Signal to Noise Ratio（SNR） is low. In response to this situation, a new algorithm is proposed. A quasi-periodic impulse signal is calculated based on the distorted signal processed by MMSE-LSA algorithm using a nonlinearity to regenerate harmonics. This artificial signal is then used to multiply the original enhanced signal, making the voice of the harmonic components to be prominent. Experimental results show that the algorithm is a good solution to the problem of Harmonic distortion better and is better than the traditional algorithm when improving the voice quality of high harmonics in low SNR.