The Research And Improvement Of The Acoustic Echo Cancellation Based On NLMS Algorithm

With the rapid development of Internet and multimedia technology, audio interaction over Internet has played an important role in the implementation of the live system(VoIP system). But acoustic echo in the real-time audio transfer severely impacts communication quality, reducing the performance of the system. How to cancel the acoustic echo in the VoIP system is the goal of the dissertation.The article summarizes the principles of the implementation of acoustic echo cancellation and basic algorithms. Based on the analysis of the characteristics of the live system, the article presents a mechanism of the echo cancellation with a novel hybrid double-talk detector (DTD) and silence detection.The flow of the proposed mechanism is described in detail. Then, the article lucubrates three key technologies: The first one is the hybrid double-talk detection algorithm, the role of double talk detector is help to freeze adaptation of the filter when near-end speech is present. The second one is the silence detection, this can help to reduce data amount in order to improve transport efficiency. The last one is the adaptive filter and its algorithm. Based on the analysis of LMS and NLMS, a variable step size LMS algorithm is adopted. Compared with traditional algorithm, it works faster and more robust.