论文部分内容阅读
该文从信息论的观点出发，对语音信号的隐马尔可夫模型（ＨＭＭ）的状态数进行研究，建立了ＨＭＭ的状态数研究的简化模型，指出ＨＭＭ的信息熵是由语音信号的固有熵和附加熵组成。随状态数增加，信息熵趋向固有熵。最后，在综合考虑信息熵和运算量两方面因素情况下，得出的状态数宜在６～８之间的结论。
In this paper, we study the state numbers of Hidden Markov Models (HMM) of speech signals from the perspective of information theory and establish a simplified model for the study of state numbers of HMMs. We point out that the entropy of HMMs is determined by the entropy of speech signal and Additional entropy composition. As the number of states increases, the entropy of information tends to be inherent entropy. Finally, considering the two factors of information entropy and computation, we conclude that the number of states should be between 6 and 8.