Research And Implementation Of Speech Enhancement In Conference Phone Based On DSP

With the teleconference system and the popularity of people's call to raise the quality requirements, the number of speakers and a microphone also called for more to enhance the immersive feeling. However, the number of speakers and microphone increasing also causes echo increasing. How to effectively eliminate these echoes has become an important factor to ensure the quality of calls.This paper mainly studies for the rapid multi-channel echo cancellation adaptive algorithm which can adapt the DSP better. Firstly, the principle and problems of the adaptive multi-channel echo cancellation are analyzed. For recently adaptive multi-channel echo cancellation algorithms and their characters are discussed systemically. Then based on the LMS algorithm using the previous data of adaptive system and block processing technology in frequency domain technology, obtain an improved adaptive echo cancellation (NM-FBLMS)algorithm and expanded this algorithm to double-channel echo cancellation. From the results of simulation ,new algorithm have better convergence rate and computational complexity previously, it obtain a faster convergence rate than the LMS algorithm and lower computational complexity than the APA and RLS algorithm. At the same time, comparing several double-talk detector algorithm, we select the optimal performance of double-threshold method for double-talk detector. Last, we carry out a real-time testing based on TMS320VC5509DSK the minimum target system board.