基于自适应滤波器的回声消除算法研究及应用-信号与信息处理专业论文.docxVIP

AbstractAbstract Abstract Abstract The echo problem exits in many fields，such as large loop delay telecommunication channels，mobile phone system，VolP,and teleconference system，etc．It is difficult to get smooth conversation without proper echo control，so to support high—speech-quality voice communication,the technology ofecho cancellation has been a very active research field in the recent years． A commonly adopted approach to deal with the problem brought by acoustic echoes in speech communication is to apply adaptive echo canceller,which involves two main issues： adaptive filtering algofithm and double-talk detection method． Firstly,this thesis conducts a detailed study on a variety of widely applied adaptive filter algorithms．Their difference ale theoretically and empirically campared．The LMS algorithm has the computational simplicity,the good robustness,but slow convergence rate．Compared to LMS，the NLMS algorithm has much faster convergence rate．Howeveg the occurrence of relevant signals slows down the convergence rate of the NLMS algorithm．The PNLMS algorithm has a very fast initial convergence rate compared to NLMS when the impulse response is sparse，and it has the annoying drawback of deteriorating quickly if the ehco path is dispersive．Compared to NLMS，the APA algorithm has much faster convergence mte，but its robustness is weaker．The RLS algorithm also has a fast initial convergence rate，but higher computation complexity．However,the common shortcoming of these algorithms is the confliction between convergence mte and steady misalignment．Here，we propose a novel echo cancellation scheme which ensembles a diverse set of adaptive filtering methods and different choices of step-size such that their advantages can be combined and improved performance can be achieved． In addition,we have also addressed the problem of double-talk detection．Compared to disturbance cancellation in principle，echo cancellation is specific case of disturbance cancellation．such that the a