一种基于双帧动态时间规整的语音识别新方法
A New Method of Speech Recognition Based on Double Frame Dynamic Time Wrapping
-
摘要: 提出了一种基于双帧动态时间规整(Double Frame Dynamic Time Wrapping)的识别方法,通过缩小帧匹配距离矩阵(Frame Match Distance Matrix)的规模从而减少计算量,提高识别速度.实验表明,基于双帧动态时间规整的识别方法在识别率有所提高的条件下,缩短了识别时间,有很好的实用价值.Abstract: To improve the recognition speed, a method of speech recognition based on Double Frame Dynamic Time Wrapping is proposed, this method decrease the size of Frame Match Distance Matrix and then increase the recognition speed. Experimental results show that the DFDTW method decrease recognition time with a better recognition rate and has practical value.
下载:
