自适应滤波器消除语音信号中混合噪声

Elimination of Mixed-noise in Speech Singal by Adaptive Filter

语音信号在实际采集和传输的过程中,往往掺杂着多种噪声干扰,比较常见的是正弦窄带干扰和高斯白噪声,而一个简单的自适应滤波器往往很难同时滤除多种噪声。为了抑制混合噪声而得到真实的语音信号,在最小均方误差(LMS)自适应算法和自适应噪声抵消原理的基础上,提出了一种两级自适应滤波器方案,第Ⅰ级在传统噪声抵消系统中加入延迟单元消除正弦窄带干扰,第Ⅱ级用LMS自适应噪声抵消器消除高斯白噪声,同时,利用Simulink模块库对所设计的两级自适应滤波器进行了建模仿真。仿真结果表明:该方案滤波器可以有效地滤除包含正弦窄带干扰和高斯白噪声的混合噪声,达到提高语音质量的目的。

The acquisition and transmission of speech signal are often mixed with a variety of noise or interference,such as sinusoidal narrow-band interference and Gaussian white noise.A simple adaptive filter is hard to filter out them simultaneously.In order to suppress mixed-noise to get a real voice signal,a two-stage adaptive filter program was proposed based on least mean square(LMS) adaptive algorithm and adaptive noise cancellation principle.By adding delay elements to a traditional adaptive noise cancellation,the first stage could eliminate the sinusoidal narrow-band interference,and the second stage was used to eliminate the Gaussian white noise.The two-stage adaptive filter was simulated by using Simulink block library.The simulation results show that the proposed filter can effectively filter out mixed-noise including sinusoidal narrow-band interference and Gaussian white noise,and improve the quality of speech signal.