多带谱相减结合感觉加权的语音增强方法研究

Speech enhancement method based on multi-band spectral subtraction plus perceptual weighting

语音增强技术是解决噪声污染的一项强有力的预处理技术,也是语音信号处理的一个热门课题。文章提出了一种多带谱相减与感觉加权相结合的语音增强方法。该方法首先对带噪语音进行多带谱相减,然后根据人的听觉掩蔽特性,对多带谱相减后的信号进行感觉加权,以便进一步降低背景噪声。仿真结果表明,该方法在较好地抑制背景噪声和残留音乐噪声的同时,保持了较好的语音可懂度。

The pronunciation-strengthened technology is a powerful treatment technology for solving the noise pollution ; it is also a hot subject in voice signal processing. In this paper a speech enhancement method based on multi-band spectral subtraction plus the perceptual weighting algorithm has been put forward, this method can suppress noise. Firstly, this method performs multi-band spectral subtraction, and then based on the acoustical masking characteristics of human auditory system a processing of weight to feeling is brought to bear upon the subtracted signals. The simulation results show that the method has excellent performance in background noise and“remnant music noise” suppression, and at the same time the speech signal keeps very low audible distortion.