Performance evaluation of an echo canceller combined with a speech coder

摘要

We accomplished performance evaluation on the echo canceller which is combined with a speech coder in this paper.The Normalized Least Mean Square(NLMS) algorithm is used to implement the adaptive digital filter that removes echoes.A combined structure is realized with the QCELP(Qualcomm Code Excited Linear Prediction) speech coder that determines the Linear Predictive Coding(LPC) coefficients and codebook information in the encoding process.utilizing these information with reasonable complexity and memory increment,echo component can be effectively eliminated and this eventually gives better convergence property than conventional algorithm.Double talk detection is done comparing current frame energy with the thresholds which are maintained based on the background noise level.The proposed algorithm is shown to have good convergence characteristics in the simulation compared to the conventional algorithm.Furthermore,the proposed echo canceller is proven to be easily combined with the QCELP speech coder.