【摘要】 提出了一种改进的变步长LMS(Modified Variable Step SizeLeast Mean Squares)算法。用误差信号的自相关e(n)e(n-1)及均方误差e2(n)来调节自适应滤波算法的步长μ(n)。仿真结果表明该算法的收敛速度更快,优于MVSSLMS算法,可以很好地应用于自适应系统中。最后,研究了改进算法在语音增强系统中的实现,从混噪语音中提取了原始语音信号。更多还原

【Abstract】 A improved variable step size algorithm is presented,which uses the estimation of the autocorrelation about error signal e(n)e(n-1) and means square error e2(n) to adjust the step of adaptive algorithm μ(n).Computer simulation result verifies the improved algorithm has faster convergence rate than MVSSLMS and can be applied to adaptive systems effectively.Finally,the technology of realization is studied in the speech enhancement systems,and retrieves clear speech.更多还原