语音识别模型hmm

语音信号分离的ICA算法。卡达索供稿

这个才是好东西，语音识别相比大家都知道，但是目前市场上出现的效果都不是太好，这个代码也是语文转换，效果还可以，但是说话不能口音太重-This one is just good thing, speech recognition compare with all know , but now, the effect that appears on market does not be too good , this code is also Chinese conversion, effect return can, but speak can not accent value too

直接对语音信号进行频谱搬移，得到转换后的语音，男声发生了变化。-Directly to the voice signal spectrum move, to be converted voice, male voice has changed.

英文数字从1到10，希望做语音识别用得上-English figures from one to ten, they hope to do useful Speech Recognition

一个简单的报时小软。可以中文报时 初学语音识别，希望同样爱好的朋友可以多多交流-A simple small-time soft. Time learning in Chinese speech recognition, hope that the same lot of loving friends can exchange

此方案基于全极点语音产生模型，编码过程简述如下：首先通过线性预测分析方法提取声道滤波器参数；其次通过合成分析法确定最佳激励矢量；最后将滤波器参数和最佳激励矢量进行编码传输。-This program is based on all-pole speech production model, the encoding process summarized as follows: First, by linear prediction analysis to extract channel filter parameters followed by analysis by synthesis method to determine the best incentive vector Finally, the filter parameters and the best incentive vector encoded transmission.

一篇关语音压缩编码的DSP实现的论文 NH文件-a customs voice coding DSP paper documents NH

混合基的傅立叶变换算法-mixed-Fourier transform algorithm

这个函数也要被melcepst函数调用，用来进行计算Mel倒谱系数。这个函数的作用是构造mel滤波器组。-This function must be melcepst function call is used to calculate Mel cepstral coefficients. The role of this function is to construct a mel filter bank.

mpeg2音频编码部分的测试原代码 需要的下载-mpeg2 audio encoding part of the testing needed to download the original code