从线性预测HMM到一种新的语音识别的混合模型

From Linear Prediction HMM to a New Combined Model for Speech Recognition

线性预测 HMM（Linear Prediction HMM,LPHMM）并没有象传统 HMM那样引人状态输出独立同分布假设,但实用中识别性能并不佳.通过分析两种HMM的各自优劣,本文提出了一种新的语音识别的混合模型,将语音静态特性（基于传统HMM）和动态特性（基于LPHMM）分别描述又有机结合在一起,更为精确地刻划了真实的语音现象,同时又继承使系统的实现改动很小和较小的计算量.汉语大词汇量非特定人连续语音识别的实验表明,混合模型的识别性能显著好于LPHMM和传统HMM.理论上,本文还给出了LPHMM的一组闭式参数重估公式.

Linear prediction (LP) HMM does not make the independent and identical distribution (IID) assumption as traditional HMM; however it often produces unsatisfactory results in practice. In this paper, a new combined model for speech recognition is proposed, based on a new analysis of both HMMs modeling strengths and weaknesses. The new model works with LPHMM as the dynamic part and traditional IID-based HMM as the static part;in addition, easy implementation and low cost are preserved. Experiments on speaker-independent continuous speech recognition demonstrated that the combined model performed much better than both LPHMM and traditional HMM.Theoretically,a new closed-form parameter re-estimation formula is suggested for training LPHMM.