基于神经网络的语音识别系统

本文研究神经网络的多层感知器模型在语音识别中的应用。为达到实时识别的效果,笔者设计了TMS32010数字信号处理板,它与PC/XT机构成主-从系统,由PC/XT机管理数据,TMS32010进行高速运算,实验中,考察了神经网络对数字语音的识别能力,识别率达到95％以上,结果表明,用神经网络进行语音识别是一种很有前途的方法。

THE SPEECH RECOGNITION SYSTEM ON THE NEURAL NETWORK

This paper describes the use of multi-layer perception model of neural network in speech recognition. In order to recognize speech in real time, we design the TMS32010 digital signal processing board, which is subordinate to a micro-computer, the PC/XT, which manages speech data, while the TMS32010 performs high speed calculation. In the experiment, the capac- ity of neural network digital speech recognition is explored, the recognition accuracy is above 95%. The experimental result shows that the speech recognition based on neural network is a kind of prospective way.