基于改进的SVSLMS算法的语音识别系统研究  

Research on speech recognition system based on improved SVSLMS algorithm

作　　者：尹秋明 沈天飞[1] 龚雪 Yin Qiuming;Shen Tianfei;Gong Xue(College of Mechanical Engineering and Automation,Shanghai Univers让y,Shanghai 201900,China)

机构地区：[1]上海大学机电工程与自动化学院,上海201900

出　　处：《电子测量技术》2020年第1期63-68,共6页Electronic Measurement Technology

摘　　要：为了克服固定步长LMS算法固有的收敛速度与稳态误差之间的矛盾,在传统SVSLMS算法的基础上提出了改进的SVSLMS算法。通过理论分析和仿真研究的方法,给出了该算法中各个参数的选取方法,并确定了适用于本文所采用的语音样本库的参数最佳值。对固定步长LMS算法、传统SVSLMS算法以及提出的改进的SVSLMS算法进行了仿真对比。仿真结果表明,提出的改进的SVSLMS算法在降噪性能、收敛速度和稳态误差这3个方面都优于其余两种算法,有效的提高了语音识别系统的准确性和稳定性。In order to overcome the contradiction between the convergence speed and steady-state error of the fixed-step LMS algorithm, an improved SVSLMS algorithm based on the traditional SVSLMS algorithm is propsed. Through the methods of theoretical analysis and simulation research, the selection methods of each parameter in the algorithm are given, and the optimal parameters of the parameters suitable for the speech sample library used are determined. The fixed step size LMS algorithm, the traditional SVSLMS algorithm and the improved SVSLMS algorithm proposed are compared. The simulation results show that the improved SVSLMS algorithm is better than the other two algorithms in terms of noise reduction performance, convergence speed and steady-state error, which effectively improves the accuracy and stability of the speech recognition system.

关 键 词：自适应消噪 SVSLMS算法 语音识别 收敛速度 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]