A Study On Acoustic Echo Cancellation And Noise Compensation Based On ADSP-BF533

This paper studied the mothod of acoustic echo cancellation based on Least Mean Square (LMS) and described an improved adaptive LMS algorithm with fast convergence speed. For the noisy environment, we presented a new echo canceller structure. This structure combined the echo cancellation and noise compensation. Based on the structure, we realized a stable echo cancellation in an embedded hardware system based on ADSP-BF533. The main work of this paper is as follow:Firstly, we presented a new serial structure for the echo cancellation. The noisy signal was compensated before the echo cancellation. A voice/noise detector based on energy threshold was realized, through it, the noise was re-estimated and updated accordingly. We gave an improved spectral subtraction algorithm to compensate the noise signal. This serial system structure improved the echo cancellation effects.Secondly, we gave an improved adaptive LMS based echo cancellation method because the conventional LMS and the Normalized LMS (NLMS) converge slowly. The improved method based on the sparse response of the echo path. During the estimation of the adaptive filter, different weights were given to the non-zero coefficients. This method speeds up the convergence process significantly.Thirdly, we designed and realized a double-talk detector. It based on the angle of real echo signal and the estimated echo signal. When there exists the near end speech signal, the adaptive filter estimates the echo signal without changing its own coefficients, but when there doesn't exist the near end speech signal, the adaptive filter not only estimates the echo signal but also changes the coefficients.Last, an ADSP-BF533 based multimedia hardware platform was developed. It includes video and audio CODEC, GPS/GPRS, network interface, etc. The audio CODEC AD1836A was extended through the SPORT interface. Based on this hardware system, we developed the low-level audio driver software and realized the acoustic echo cancellation algorithms in the embedded application environment. The system achieved good performance.This paper presented an improved realization of the acoustic echo cancellation and noise compensation. An ADSP BF533 based hardware system platform was developed. An embedded acoustic echo canceller was developed on this hardware platform and achieved good effects.