English

应用科学学报 ›› 1995, Vol. 13 ›› Issue (3): 314-318.

• 论文 • 上一篇    下一篇

MHMM和ANN法结合的语音识别方法

胡光锐, 周浩, 严永红   

  1. 上海交通大学
  • 收稿日期:1994-04-25 出版日期:1995-09-30 发布日期:1995-09-30
  • 基金资助:
    国家攀登计划认知科学(经神网络)重大关键项目的资助

A SPEECH RECOGNITION SYSEM BASED ON MHMM AND ANN METHOD

HU GUANGRUI, ZHOU HAO, YAN YONGHONG   

  1. Shanghai Jiao Tong University
  • Received:1994-04-25 Online:1995-09-30 Published:1995-09-30

PDF

141

被引次数

6

摘要： 提出了一种改进的隐马尔柯夫模型(MHMM)方法,并在MHMM方法中结合进人工神经网络(ANN)方法,以这种方法为基础建立了语音识别系统。实验结果表明,与隐马尔柯夫模型方法比较,新系统的识别率有所提高,从而,充分地显示了人工神经网络在语音识别中的可行性和有效性。

关键词: 信号处理, 语音处理, 语音识别

Abstract: In this paper, we present an improved hidden Markov model (MHMM) method,and propose to incorporate the artificial neural network (ANN) method in the MHMM framework to set up a speech recognition system.We experimentally show that MHMM has been successfully combined with a neural network to build a new system.It is found that the recognition performance of the new system is much better than that of the conventional HMM recognizer. The new system also demonstrates the usefulness of ANN in speech recognition.

Key words: speech recognition, speech processing, signal processing