遗传算法与神经网络在语音识别中的应用

作者：刘俊华1,颜运昌2,荆 琦2 日期：2007-12-28/span> 浏览：3830 查看PDF文档

遗传算法与神经网络在语音识别中的应用
刘俊华1,颜运昌2,荆 琦2
（1.中国空空导弹研究院，河南 洛阳 471009; 2.湖南大学 机械与汽车工程学院，湖南 长沙 410082）

摘 要：为克服隐马尔可夫模型(HMM)在语音识别应用中的缺点和局限性，在语音识别中引入神经网络。采用前馈多层神经网络，用遗传算法和BP算法结合的方式对网络进行训练，克服了传统BP算法收敛速度慢、易陷入局部最小等缺陷。仿真实验结果表明，该模型有效地提高了语音的识别率，并缩短了识别时间，实现了效率与时间的双赢，为算法的实用性奠定了基础。
关键词：马尔可夫模型；语音识别；神经网络；遗传算法

Application of genetic algorithm and neural network in speech recogition
LIU Junhua1, YAN Yunchang2, JING Qi2
(1.China Airborne Missile Academy, Luoyang 471009, China;
2.College of Mechanical & Automobile Engineering, Hunan University, Changsha 410082, China)

Abstract: To overcome for these weaknesses of hidden markov models (HMM) used in speech recognition, neural networks has been introduced in speech recognition. An approach using multilayer perceptron (MLP) and training it by combining genetic algorithm (GA) with BP algorithm was proposed. Experiment results show that the training speed can be accelerated by the method, and the recognition performance is also promoted.
Key words: hidden markov models (HMM); speech recognition; nueral network (NN); genetic algorithm (GA)