基于分类识别的隐马尔可夫模型语音识别技术

通过分析基于隐马尔可夫模型(HMM)语音识别的原理,针对模板提取过程中语音信号的基音频率差别增大而出现的语音识别率下降的问题,提出分类识别的方法,通过采用基音周期(Pitch)判决方法,将特征相近的帧合并,并计算基音频率的MEL频率倒谱系数,采用隐马尔可夫模型(HMM)进行语音识别,最终通过仿真实验验证分类识别方法对语音识别率提高的影响,得出此方法的适用环境和范围.

Hidden Markov Models for Speech Recognition Technology Based on Classification and Identification

The principle of speech recognition based on Hidden Markov model（ HMM） is analyzed. According to the decline of speech recognition rate,which is caused by pitch frequency difference increases of the speech signal in the processing of template extraction,the method of classification and recognition is put forward.Using pitch detection the frames with the similar characteristics are merged,and the MEL frequency cepstrum coefficients of pitch frequency is calculated. Then the method of Hidden Markov model（ HMM） is used to speech recognition,the influence of the speech recognition rate through the simulation experiment is test. The applicable environment and scope are obtained finally.