Open source for Automatic Speech Matching ?

 

Automatic Speech Matching is not Automatic Speech Recognition, which is to compare two pieces of speech audio signal and return how many percentages these two audio signal match.

This tech will usually be used like such scenarios:

  • Pronounce learning, for example, there is standard pronounce signal of word "Hello", for students who are learning English, they pronounce their own "Hello", so we need use the ASM tech to compare how similar or how correct the student pronounced. So we need figure out sort of algorithm to compare these two 1-D audio signal.

    • 2.We can extend from above from single word to a sentence. Then how to match these audio signal ?

    Question here is to look for some good open source or commercial solution for ASM.

    Or any other good solutions for such real requirements ?

    Thanks in advance !

    The comparison with the template will not give anything good because it will not actually hint what was spoken incorrectly. The good pronunciation learning framework doesn't match with the template but with the acoustic model representing proper and wrong pronunciation. This way it can detect errors in speech which learner make. You can read

    The SRI EduSpeakTM System: Recognition and Pronunciation Scoring http://www.speech.sri.com/people/hef/papers/EduSpeak.ps

    For implementation of this algorithm on IPhone you can check

    http://ottercall.com

    链接地址: http://www.djcxy.com/p/34430.html

    上一篇: 目前语音识别的一般状态是什么？

    下一篇: 自动语音匹配的开放源代码？