论文部分内容阅读
对于目前在语音识别中广泛使用的两种技术即动态时间规整（DTW）技术和隐马尔可夫模型（HMM）的本质联系，提出了二者的统一模型（DHUM，DTWandHMMUni-fiedModel），并分别给出DTW和HMM向DHUM的转换关系。文中还提出了用DHUM解决更接近语音实际情况的高阶HMM作语音识别时所面临的运算量过大的问题。中等词表的识别实验结果表明，建立在DHUM之上的识别器的识别性能不低于DTW和HMM识别器。
The two models (DHUM, DTWandHMMUni-fiedModel) are put forward for the two technologies widely used in speech recognition, namely the dynamic time warping (DTW) technology and the hidden Markov model (HMM) The conversion between DTW and HMM to DHUM is given. The paper also raised the problem of using DHUM to solve the problem of too much computation in high-order HMM for speech recognition that is closer to the actual situation of speech. The experimental results of middle vocabulary recognition show that the recognition performance of recognizers based on DHUM is not lower than that of DTW and HMM recognizers.