BP神经网络应用于孤立词语发音识别的研究

ON APPLYING BP NEURAL NETWORK TO ISOLATED WORDS SPEECH RECOGNITION

介绍了BP神经网络的学习规则和用于语音识别的基本原理,建立了一个用于常用孤立词语音识别的BP神经网络,选择声道反射系数为语音识别的特征值,建立了网络的训练样本集,对网络进行了训练;用MATLAB进行了识别仿真,表明能较好地实现孤立词语音识别。

The learning principle of BP Artificial Neural Network （ANN） and the basic theory on speech recognition is introduced in this article. A BP Artificial Neural Network for common isolated word recognition is presented. Selecting reflection coefficients of vocal track as feature value of the speech recognition to construct the training sample set of BP Artificial Neural Network, the ANN is trained. The recognition results obtained from MATLAB simulation show that the speech recognition of isolated words achieves well.