Research On Algorithms Of Adaptive Acoustic Echo Cancellation

Adaptive acoustic echo cancellation is one of the key technologies for the modern telephone conferencing systems and hands-free mobile phone calls devices, which can provide full-duplex telecommunication.Acoustic Echo directly or through more times reflection which together inputting the microphone, leading to the echo reverberation time is very long. In order to achieve echo cancellation, adaptive filter may be more than 1,000 coefficients. So, use this adaptive filter to access the Echo model, the computation so complexly; Because the voice signal strong correlation, resulting in slow convergence .Adaptive algorithm can not change when the echo channel property has been changed, the effect of the echo cancellation relatively poor; During the double-talk, the near-end speech signal will serve as energy proximal lot of noise into the adaptive filter, which could lead to adaptive filter coefficients divergence; Meanwhile, stereo acoustic system can provide more real conversation environment, but also brings new problems. This paper will analysis the adaptive filter echo cancellation of the problems still exist ,to propose the improved acoustic echo cancellation adaptive algorithms. This paper work and research results are as follows:1. Against the room impulse response a long time, the complexity computation and the slow process of convergence problem of the adaptive filtering algorithm , this paper proposed an improved over-sample delayless sub-band dual-channel echo canceling algorithm, the algorithm can maintain the sub-band method saving the computation and improving the convergence. At the same time, through mapping the sub-band adaptive filter to the full band filter avoid signal delay and adopting the double-channel filters make the echo canceling system working under the least residual echo power.2. Against the adaptive filter weights in the double-talk conversation becoming divergent, and an additional double-talk detector not only costly, but also freezing the filter coefficients of the filter may have deviated from the optimal solutions, which lead to the echo cancellation performance deterioration, this paper proposed a time-varying power spectrum no double-talk detection algorithm. During the double-talk conditions, no need to double-talk detection of the state also can echo cancellation.3. This paper will continue to research the stereo echo cancellation system based on the monaural system. Analysis the routine lineally combined adaptive filter can not correct identify the echo impulse response, then propose the improvement echo canceling method. The simulation results show the improved stereo echo cancellation converged property.