基于小波神经网络的嵌入式语音识别系统

Embedded Speech Recognition System Based on Wavelet Neural Network

以ARM和嵌入式操作系统Windows CE为平台设计了基于小波神经网络的语音识别系统。首先介绍了语音识别的基本原理、系统的硬件组成、系统的功能结构、系统的工作原理,然后叙述了小波和神经网络的理论知识及算法,最后介绍了嵌入式语音系统的软件设计。由于采用S3C2410芯片负责系统的控制和识别工作,而语音信号的特征提取和训练算法全部由SPCE061A芯片来完成,从而使得语音识别系统识别率提高,可移植性好。

With ARM and embedded operating system Windows CE as the platforms,a embedded speech recognition system based on wavelet neural network is designed.This paper first gives the basic principle of speech recognition,the system hardware,the system functional structure,and the working principle,then it describes the theoretical knowledge and algorithm of wavelet neural network,and finally discusses software design of the embedded speech recognition system.With chip S3C2410 for control and recognition,the extraction of voice signal feature and the training algorithm are completed by SPCE061A chip.Thus the system is of high recognition probability and good transplantability.