论文部分内容阅读
分析了由于说话人受到重力加速度变化而产生的变异语音(应力影响下的变异语音)的特点,把变异语音分成主体部分和变异部分两方面进行研究,提出一种动态时间规正与差别子空间相结合的孤立词语音识别方法。该方法构造差别子空间去除变异部分的影响,利用语音的主体部分进行识别,采用动态时间规正技术对语音特征矢量进行长度对齐,并提出了相应的判别标准。实验结果表明,本方法对应力影响下的变异语音具有良好识别效果,对15个词的小词表,系统平均识别率达到98.3%,与正常语音的识别率基本相当。克服了话者在应力影响下由于心理紧张和生理情况的变化,语音发生严重变异,导致常规语音识别系统性能急剧下降的缺点。
This paper analyzes the characteristics of the variant speech (variation speech under the influence of stress) caused by the change of the speaker’s accelerometer, divides the variation speech into the main part and the variation part, and proposes a dynamic time alignment and difference subspace A combination of isolated words speech recognition method. The method constructs the differential subspace to remove the influence of the variation part and uses the main part of the speech to recognize the length of the voice feature vector using the dynamic time alignment technique and puts forward the corresponding criterion. Experimental results show that this method has a good recognition effect on variant speech under the influence of stress. The average recognition rate of the system is 98.3% for the 15-word small vocabulary, which is almost the same as that of normal speech. Overcomes the speaker under the influence of stress due to psychological stress and physiological changes, a serious voice variation, leading to the performance of conventional speech recognition system, the sharp decline in the shortcomings.