论文部分内容阅读
在传统的基于隐马尔科夫模型的语音口令识别系统的误识口令中,有很大一部分是将字数为两个字的语音口令误识为字数大于两个字的语音口令.针对此问题,本文提出了一种二字语音口令字数确认方法,分别以短时能量、Teager能量算子和信息熵作为确认二字语音口令的三种特征,运用滑动窗确认测试语音口令是否为二字语音口令,并根据确认结果对语音口令识别系统的识别结果进行再处理,减少了二字语音口令的误识率.在加入了字数判别功能的改进语音口令识别系统上的实验结果表明,本文提出的算法有效地提高了基于隐马尔科夫模型的语音口令识别系统的识别率.
In the traditional password recognition system based on Hidden Markov Model, a large part of the misrecognized password is misrecognized as a two-word speech password with a word count of two words.In response to this problem, In this paper, we propose a method to confirm the number of words in two-word speech. Three functions of short-time energy, Teager energy operator and information entropy are respectively used to confirm two-word voice password. The sliding window is used to confirm whether the test voice password is a two-word voice password , And then re-processes the recognition result of the speech password recognition system according to the confirmation result to reduce the misrecognition rate of the two-word speech password.Experimental results on the improved speech password recognition system with word number discrimination show that the proposed algorithm Effectively improves the recognition rate of voice password recognition system based on Hidden Markov Model.