基于解相关变步长的改进型语音增强算法

Modified speech enhancement algorithm based on de-correlation variable step size

复杂环境中噪声干扰严重影响语音信号的质量,无法正确传达语义,因此语音增强处理十分必要。传统语音增强技术存在适应性差、输入信号高度相关时收敛速度慢等问题。综合变步长最小均方(VSSLMS)算法与解相关的优点,提出了一种改进的语音增强算法,优化自适应滤波算法中步长的大小和权矢量的更新方向,提高语音降噪收敛速度。同时算法引入了连续块处理理论归一化权矢量,以提高其在嵌入式系统实现上的稳定性。仿真测试表明该算法收敛速度快、跟踪性能强,能有效去除强噪语音信号中的噪声,提高语音的清晰度与可懂度。

Noise seriously affects the quality of speech signal under complex environment, leading us cannot convey the semantic correctly, so speech enhancement processing becomes very necessary. Traditional technology exist problems as follows： poor adaptability, slow convergence speed when the input signals are heavily correlated. Therefore, an improved algorithm unifies the advantages of Variable Step Size Least Mean Square （VSSLMS） algorithm and de-correlation was proposed to increase the convergence speed by optimizing the step size and update direction of the weight vector in adaptive filter. In order to improve the stability of algorithm implementation in embedded systems, continuous block processing principle was introduced to normalize the weight vector. Simulation tests show that the novel algorithm has fast convergence speed and good performance of tracing time-varying signals. The noise can be removed effectively from speech signal with strong noise, and the speech definition and intelligibility are improved significantly.