SVAC音频编码的特征参数量化器改进  

Analysis and improvement of characteristic parameters quantizer in SVAC surveillance audio coder

作　　者：舒若 李世宝[1] 潘辛 

机构地区：[1]中国石油大学(华东)计算机与通信工程学院,山东青岛266580

出　　处：《信息技术》2014年第6期50-54,共5页Information Technology

基　　金：中央高校基本科研业务费专项资金资助项目(12CX04078A;12CX04077A)

摘　　要：SVAC监控音频编码器在前端对提取的特征参数量化编码,以防止语音编码失真影响说话人识别。特征参数编码占用码率较低,因此对其量化器有一定的技术要求。分析了SVAC的特征参数量化器编码性能,对多种情况下的量化失真进行了实验对比,根据结果有针对性地设计了新码本。In order to prevent the influence of speech coding distortion on the speaker＇ s recognition, the extracted characteristic parameters are quantified and coded in front-end in SVAC surveillance audio encoder. The characteristic parameters take up low bit rates, so the quantizer should meet certain technique requirements. This paper analyzes the coding performance of the characteristic parameter quantizer in SVAC, and makes comparison of quantization distortion in many cases. At last, new codebooks are designed according to the tests.

关 键 词：监控音频编码 梅尔频率倒谱系数 分裂矢量量化 LBG算法 

分 类 号：TN912.32[电子电信—通信与信息系统]