PID控制原理在语音信号盲源分离中的应用

语音信号在非平稳系统中是动态混合的，为了实时抑制盲源分离过程中的非平稳混合扰动，加快收敛速度，减小稳态误差，提出了一种应用PID控制原理的自适应盲源分离算法。依据一种无预处理的自适应盲源分离算法建立PID控制模型，调节学习速率，跟踪语音信号的分离过程，实时减小由非平稳混合引入的分离误差，动态更新分离矩阵。在混合矩阵缓变和突变两种情形下分别对PID参数整定和语音信号的分离进行仿真分析，结合经典算法对比提出算法的性能。仿真与对比结果表明，提出的算法适用于非平稳混合系统语音信号的分离，算法性能较经典算法有改善。

Application of PID Control Theory in Blind Source Separation for Speech Signals

Speech signals are mixed dynamically in non-stationary systems.In order to suppress non-stationary mixing disturbances real-timely in blind source separation process,accelerate convergence speed and decrease steady-state error,a self-adaptive blind source separation algorithm based on PID control theory is proposed.According to the self-adaptive blind source separation algorithm without pre-processing,the PID control model was established.Learning rates were regulated to trace the separation process of the speech signals,decrease the separation error real-timely caused by non-stationary mixture and update the demixing matrix dynamically.PID parameters tuning and speech signals separation were simulated under mixing matrix changing slowly and abruptly,and its performance was contrasted with the classical algorithm.The results show that the proposed algorithm is applicable to the separation of the speech signals in non-stationary mixing system and its performance is improved.