Variable-length coding of ACELP gain using Entropy-Constrained VQ

We designed variable length code for the vector quantization (VQ) gain parameter of the Algebraic Code-Excited Linear Prediction (ACELP) speech coding scheme, aiming at reduction of bit rate and distortion in the environment of IP communication. The code index is selected taking both quantization distortion and average code length into consideration. The VQ tables has been trained by the algorithm based on Entropy-Constrained VQ (ECVQ). It has been shown that this scheme can keep the quality and reduce the average bit rate by 0.2 kbit/s when applied to the state-of-the-art speech coding standard ITU-T G.718.