基于短时分数阶傅里叶变换的语音增强算法

Method for Speech Enhancement Based on Short-Time Fractional Fourier Transform

提出了一种基于短时分数阶傅里叶变换(STFRFT)的语音增强新方法。该方法首先将带噪语音信号进行短时分数阶傅里叶变换,然后在分数阶傅里叶域(FRFD)对信号进行滤波,最后对滤波后的信号进行短时分数阶傅里叶逆变换,得到增强后的语音信号。实验表明在选定最佳的分数阶阶数时,可使噪声得到最大限度的滤除,大大提高了语音增强效果。

A new method for speech enhancement is proposed,which is based on short-time fractional Fourier transform（STFRFT）.Firstly,noisy signal is processed by STFRFT.Then,the noisy signal is filtered in fractional Fourier domain（FRFD）.Finally,the enhanced signal is obtained by using inverse short-time fractional Fourier transform（ISTFRFT） to the filtered signal.Simulation experiments indicate that when it is given an optimal fractional order,noise can be denoised maximumly,which greatly improves the quality of speech enhancement.