基于谐波重建的语音增强算法的研究

Speech enhancement using adaptive gain function based on harmonic recovering

目前,对于一些信号处理方面的应用,包括移动通信、语音识别和助听设备等,语音增强都是一项重要的任务。针对在频域上语音信号的大部分能量集中于浊音部分的谐波结构这一特点,文中提出了一种基于谐波结构重建的语音增强方法,通过从频域上尽可能地恢复语音的谐波结构并抑制信号中的噪声成分来达到语音增强的目的。其中谐波结构提取部分采用改进的频域峰值选取算法和谐波再处理策略来准确定位频域上浊音部分的谐波结构的位置。根据语音频谱中谐波结构的分布特性,文中采用梳状滤波器来设计增益函数,并利用噪声估计方法设计频域增益函数的增益值。对于带噪语音,该方法能有效地恢复其谐波结构并抑制语音中的噪声成分。此外文中还将此方法与几种经典的语音增强方法针对两种噪声进行比较实验,实验结果表明该方法的性能整体上优于其他几种语音增强算法,尤其是在低信噪比条件下文中所提出的方法性能优势明显。

Speech enhancement is an important task for many signal processing applications,including speech recognition,mobile communications and hearing aids. This paper proposes an enhancement method for single-channel speech signals based on harmonics recovering. Given the fact that most of the energy of speech signal is concentrating on harmonic structures of voiced frames,the paper puts forward a strategy to precisely locate the harmonic structures of voiced frame in frequency domain. And then noise estimation is used for spectral gain calculation of the spectral gain function. It can retrieve the harmonic structures of voiced speech and suppressing the noise component in the frequency domain. The performance of the proposed speech enhancement method is evaluated by several quality tests under white Gaussian noise（ WGN） and babble noise environment. According to the experimental results,the proposed enhancement method achieves better performance than the other three existing methods as a whole,particularly at low signal-noise-ratio（ SNR） conditions.