摘要
介绍了一种小词汇量的语音识别方法。在传统的DTW语音识别的基础上,通过放宽端点、对路径进行斜率限制和改进局部路径限制等方法,使系统能够实现快速的识别响应,尤其是对于小词汇量和特定人识别,可以有效地提高响应速度和识别率。
A fast speech recognition system for small vocabulary is presented. Based on traditional DTW speech recognition and through relaxing extreme, limiting route＇s slope and improving some routes, the system can recognize and respond quickly especially when applied in small vocabulary speech recognition and speaker dependent recognition.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2006年第2期10-13,共4页
Journal of Wuhan University of Technology：Information & Management Engineering
关键词
语音识别
DTW
线性预测
端点检测
speech recognition
DTW
linear prediction
endpoint detection
