嵌入式数字语音拨号系统的研究与实现

将语音识别技术应用于拨号系统，在嵌入式平台上实现了一款针对非特定人的数字语音拨号系统。语音识别算法中选择梅尔频率倒谱系数为特征参数，连续隐马尔科夫模型。为训练和识别过程模型，利用Qt界面对识别过程进行控制，系统针对非特定人数字语音识别进行实验。结果表明，系统针对非特定人识别率达到了98％，识别时间为3．55S。识别率和实时性都满足语音拨号的需求。

Research and Implementation of Speaker - Independent Digital Voice Dialing System

Speech recognition technology is put into practice, a speaker- independent digital speech dial system is intro- duced. In the speech recognition process, the MFCC is used as the feature parameter and the CHMM is chosen as the train- ing and recognition model. The recognition by Qt GUI is controled and this system on speaker - independent digital recogni- tion is tested. The result indicates the accuracy of this system reaches up to 98% to speaker - independent ease and it costs 3.55s. It can meet both the accuracy and time requirements of speech recognition system.