HTK工具-提特征
作者：互联网
- 可使用Hcopy来提特征。以16k语音提取mfcc特征为例，指令如下：
HCopy -T 1 -C config -S codetr.scp
- config如下：
SOURCEKIND = WAVEFORM
SOURCEORMAT = WAV
SOURCERATE = 625
WINDOWSIZE = 250000
TARGETRATE = 100000
TARGETKIND = MFCC
SAVEWITHCRC = T
ZMEANSOURCE = T
USEHAMMING = T
PREEMCOEF = 0.97
NUMCHANS = 26
NUMCEPS = 12
CEPLIFTER = 22
- 配置项说明：
SOURCEKIND = WAVEFORM
SOURCEORMAT = WAV
SOURCERATE = 625 #一个采样点对应的时长，单位100ns
WINDOWSIZE = 250000 #窗长25ms
TARGETRATE = 100000 #窗移10ms
TARGETKIND = MFCC #特征类型
SAVEWITHCRC = T
ZMEANSOURCE = T #去直流分量（过零检测）
USEHAMMING = T #加汉明窗
PREEMCOEF = 0.97 #预加重系数
NUMCHANS = 26 #滤波器组的个数
NUMCEPS = 12 #倒谱参数的个数（做DCT后保留的倒谱个数）
CEPLIFTER = 22 #倒谱提升系数（对高频的22个参数进行加权）
- codetr.scp如下：
/root/sjy/waves/S0001.wav /root/sjy/train/S0001.mfc
/root/sjy/waves/S0002.wav /root/sjy/train/S0002.mfc
/root/sjy/waves/S0003.wav /root/sjy/train/S0003.mfc
/root/sjy/waves/S0004.wav /root/sjy/train/S0004.mfc
- 说明：
每一行的左边为待提特征的音频，右边为特征的保存路径
标签：sjy,特征,HTK,mfc,train,waves,wav,工具,root 来源： https://blog.csdn.net/zhou961413764/article/details/101219309