基于短时谱估计的语音增强改进算法

Improved Algorithm of Speech Enhancement Based on Short- time Spectrum Estimation

语音在传输过程中受到来自周围环境、传输媒介等的干扰是不可避免的,这些干扰会严重影响语音接收时的质量,导致收到的语音信号不再是原始的纯净语音信号,而是带有各种干扰噪声的语音信号,这不仅影响语音的收听质量,也给后续的语音处理带来了一定的影响。因此对语音进行增强不可或缺。大部分传统的语音增强算法仅仅只通过改变语音的幅度,再叠加上原始的语音相位或者仅调整语音的相位再和未改变的幅度叠加来实现语音信号重建从而增强语音。本文提出了一个通过既改变语音信号的幅度又改变其相位的语音增强算法。通过使用客观语音质量测评(PESQ)和语谱图对用不同方法增强后的语音进行比较,验证了用本文方法得到的增强语音质量更佳。

Speech signals will be disturbed inevitably by environmental factors and transmission media during transmission. It leads to the lower quality of received speech,and the speech turns into noisy speech rather than the original clear speech,which can not only influence the voice receiving quality but also the post processing of speech. Hence,speech enhancement is very essential. Typical speech enhancement algorithms only modify the magnitude spectrum recombined with the unchanged phase spectrum or adjust the phase spectrum recombined with unchanged magnitude spectrum to reconstruct the enhanced speech signal. In this paper,a new method was proposed,which enhanced the speech by the way to change both magnitude spectrum and phase spectrum to get reconstructed speech that enhanced. The objective speech measure PESQ test and spectrogram analysis had proved that the proposed method earns outperformance among the typical algorithms.