欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
目前语音识别技术已被广泛应用于嵌入式手持设备当中，而端点检测作为语音识别系统的前端处理模块发挥着重要的作用。传统端点检测的方法是基于固定门限阈值，所以不能够随时根据语音环境的变化，准确检测出端点的实际位置。利用短时平均上升过零间隔和短时平均幅度差作为门限参数，采用自适应更新门限阈值的方法，获得了优于固定门限阈值的检测结果。同时叙述了利用TI推出的OMAP5912嵌入式多媒体应用平台实现算法开发的过程。
英文摘要：
Nowadays speech recognition technologies have been applied in the embedded handheld device widely. It is especially critical to detect the endpoint of signals accurately for speech recognition system. The traditional way is based on the invariable threshold which has intrinsically limited to vary from the changing of the environment. So short time rise zero-crossing interval and magnitude difference are taken as arguments to renew the thresholds. Good test results are obtained with the adaptive threshold than traditional ways. This paper discusses the processes to accomplish algorithms by using TI＇ s OMAP5912 embedded multimedia platform.
同期刊论文项目
同项目期刊论文
期刊信息
位置：