【作者】 丁琦； 徐金甫； 王炳锡；

【Author】 Ding Qi Xu Wang Xu Jinfu Information Engineering University,No.304,P.O.Box 1030,Zhengzhou,Henan,P.R.China

【机构】 信息工程大学；

【摘要】 本文提出了一种基于MELP编码器的速率可变的极低速率语音编码算法。该算法采用超级帧结构和参数联合矢量量化技术降低语音信号帧间的冗余度,利用超级帧中各帧线性预测滤波器的相似性,使用变速率语音编码技术降低平均比特率。文中定义了一个新的LSF距离测度,用来度量两组不同的LSF参数所表征的滤波器间的差别,并由距离测度决定每超级帧的速率。对算法解码语音质量的客观测试表明本文提出的语音编码算法可以在所使用的560bps和800bps两种速率下保持可接受的语音质量。更多还原

【Abstract】 This paper proposes a novel variable low bit rate speech coding algorithm based on Mixed Excitation Linear Predictive (MELP) coder.This coder use a superframe structure and joint quantization methods to reduce the inter-frame redundancy of the parameters.In each superframe,the similarity of the linear prediction filter denoted by LSF parameters is imposed to reduce the bit rate by variable rate speech coding.A new distance measurement for LSF is proposed to describe the difference between two sets of LSF parameter.And the coding rate is decided by the distance measures of different LSF parameters in every superframe.Objective test results of synthesized speech quality show that the proposed speech coder is capable of maintaining acceptable speech quality at the rates 560 bps and 800 bps.更多还原