摘要
利用话音信号在进行信源编码后数据流中的特征,可以识别数字话音数据流的编码样式和参数。针对大量话音码流样本数据,利用统计分析方法给出自适应多速率(Adaptive Multi Rate,AMR)语音编码的背景噪声帧、零一分布、游程分布及参数分布等特征,在此基础上提出了基于学习的AMR语音编码识别方法,并给出AMR语音编码的特征模型,通过计算机仿真比较了话音帧分析法和统计特征分析法在不同误码条件下的识别性能。
The coding pattern and parameters of digital speech data stream can be recognized based on the characteristics of speech coding stream,which is generated by coding the analog speech signals.Aiming at large amount of speech stream sample data,such characteristics of AMR speech coding as background noise frame,probability of bit one and zero,run path distribution and parameters distribution,etc.are presented by using statistical analysis method.Based on it,a learning-based recognition method of AMR speech coding is proposed,and the characteristic model of AMR speech coding is given.Finally,the recognition performances of speech frame analysis method and statistical characteristics analysis method are compared under different bit error conditions by computer simulation.
出处
《无线电工程》
2013年第8期54-57,共4页
Radio Engineering
关键词
AMR
码流特征
识别算法
AMR
coding stream characteristics
recognition