ANN/HMM混合模型在语音识别中的应用

ANN/HMM Hybrid Model in Speech Recognition

结合HMM较强的处理时间序列的能力以及ANN的学习能力强、识别速度快等特点提出了一种ANN/HMM混合模型,该模型具有较强的处理时间序列的能力。本文主要介绍了该模型的结构以及模型的训练算法。在此基础上将其应用于语音识别的建模,并通过相应实验验证了该模型的可行性。

This paper introduces the ANN/HMM hybrid model and its application to speech recognition. The hybrid model combines the ability of HMM to deal with the time series and the learning ability of ANN. The training algorithm of the model is also presented.