Research On Adaptive Multi-Channel Acoustic Echo Cancellation Theory And Its Applicaiton

Echo in telephones is generally undesirable and inevitable. With the wide application ofteleconference and hand-free systems, the acoustic echo cancellation becomes heat-topic atabroad. However the research results of multi-channel acoustic echo cancellation are much lesscompared to those of mono-channel. There are many theories, approaches and implementationtechniques to be studied. In this thesis we achieve the following for multi-channel echocancellation.1. A new method is proposed where fuzzy logic is applied to the nonlinear parameter inthe pre-processing of multi-channel echo cancellation. Assuring the communicationquality farthest, the method can quicken the convergence of the filters at the cost oflittle computation。2. A new NLMS-like algorithm is proposed which can protect the adaptive filterperformance during double talk without setting a detector. Because the variety of thecorrelation of the far-end signal and near-end received signal can judge the double talkoccurring or echo path change ,the correlation coefficient is directly used in theiteration formula to control the iteration speed of the filter. The new algorithm has thegood double talk protection performance and also has the fast tracking performance asthe echo path changes but with less computational complexity contrast to the congeneralgorithm, and is easy to realize.3. A new multi-channel echo cancellation structure is proposed where a simplified blindsignal separation model is integrated into the normal stereophonic acoustic echocancellation. The integrated new structure translates the stereophonic echocancellation problem into two single-channel echo cancellation problems so as toavoid the non-uniqueness problem and doesn't influence the audio quality.4. Studying the method of synthesizing the stereo signals from remote mono signalsusing simplified head related transfer function and how to simplify the computation.At the same time, the localization and what is need to improve also is pointed out.Suggestion about synthesizing methods aiming at different application situations andpromising echo cancellation structure is presented.