基于GSMEFR编码的语音识别系统

主要讨论了对于小词汇量孤立词的汉语语音识别问题。文中采用欧洲通信标准委员会(ETSI)发布的GSMEFR声码器对输入语音信号进行特征提取。在识别模块采用基于动态规划原理的动态时转弯析(DTW)技术对提取的测试语音参数与训练模板的语音参数进行匹配，得到的误差最小值即为识别结果。实验结果表明，对于20条语音命令，识别准确率达到96％以上，识别时间小于1．5s。

Voice Recognition System Based on GSM EFR Code

This paper discusses the Chinese voice recognition system with small vocabulary size and isolated words. Adopting GSM EFR sounder published by ETSI, this paper makes the characteristic extraction of the input voice signal. By using DTW technique based on the theory of dynamic programming in recognition platen to match these extracted test voice parameters with the voice parameters of training platen, the minimum error value, which is the result of recognizing, is obtained. The test result denotes that the recognition accuracy can reach 96% and the recognition time is less than 1.5 seconds for 20 items of voice demands.