一种基于信号子空间和听觉掩蔽效应的语音增强方法

Speech Enhancement Based on Signal Subspace and Hearing Masking Properties

提文推导了基于离散余弦变换(DCT)的子空间分解法对有色噪声背景下的语音进行增强的公式,用基于听觉掩蔽效应的感智滤波器对增强后的信号频谱进行平滑以抑制背景噪声。几种噪声背景下对增强语音的客观测试表明,本文提出的方法可以有效地减少语音信号的失真度。

In this paper, speech enhancement in colored noise is achieved by signal subspace approach based on discrete cosine transform. A perceptual filter based on baring masking properties is employed to smooth the enhanced speech spectral and to suppress the musical noise. The algorithm is tested in different noise background. Subjective measure demonstrates that this method can reduce the distortion to the speech signal.