摘要
端点检测是语音信号处理过程中非常重要的一步,它的准确性直接影响语音信号处理的速度和结果,因此端点检测方法的研究,特别是在噪声环境下端点检测的研究,一直是语音信号处理中的热点。从基于时域参数、频域参数、时频参数、模型匹配等方法的角度,较全面地回顾了端点检测方法的发展历程,对各种方法的优缺点进行了比较分析,并给出了这些方法的改进意见,对端点检测未来的研究方向进行了展望。
Endpoint detection, whose accuracy directly affects the speed and the results, is a very important step for speech signals processing. The research of endpoint detection algorithm is a hotspot of speech signals processing, especially in adverse environment. This paper introduced different methods based on time domain, frequency domain, time-frequency domain parameters and model matching algorithms. Meanwhile, it analyzed the advantages, drawbacks and improvements of those algorithms. It also discussed the prospects of endpoint detection.
出处
《计算机应用研究》
CSCD
北大核心
2008年第8期2278-2283,共6页
Application Research of Computers
基金
上海市教委青年基金资助项目(04AB72)
上海市启明星计划资助项目(04QMX1441)
关键词
语音信号处理
端点检测
鲁棒性
speech signals processing
endpoint detection
robustness