计算机与现代化 ›› 2011, Vol. 1 ›› Issue (8): 29-31.doi: 10.3969/j.issn.1006-2475.2011.08.009
• 人工智能 • 上一篇 下一篇
李灵光
收稿日期:
修回日期:
出版日期:
发布日期:
LI Ling-guang
Received:
Revised:
Online:
Published:
摘要： 端点检测是语音识别中的一项关键技术，检测的准确性对语音识别的性能有很大的影响。本文提出一种时频结合的基于子带谱熵和短时能量的端点检测方法，和常用的双门限法、谱熵法相比在低信噪比下检测的准确率有很大提高。该方法具有抗噪性、实时性等特点，适用于实际的语音识别应用。
关键词: 端点检测, 子带谱熵, 时频结合, 短时能量, 信噪比
Abstract: Endpoint detection is an essential technology in a speech recognition system, its accuracy has a very tremendous influence on the speech recognition performance. This paper proposes a band-partitioning spectral entropy and short-time energy based endpoint detection algorithm based on time-frequency joint. Compared with the two-door method and spectral entropy, its precision is greatly improved. This method has characteristics for noise immunity and real-time, and can be applied to practical application of speech recognition.
Key words: endpoint detection, band-partitioning spectral entropy, time-frequency joint, short-time energy, SNR
中图分类号:
TP391
李灵光. 一种时频结合的抗噪性端点检测算法[J]. 计算机与现代化, 2011, 1(8): 29-31.
LI Ling-guang. Anti-noise Endpoint Detection Algorithm Based on Time-frequency Joint[J]. Computer and Modernization, 2011, 1(8): 29-31.
0 / / 推荐
导出引用管理器 EndNote|Ris|BibTeX
链接本文: http://www.c-a-m.org.cn/CN/10.3969/j.issn.1006-2475.2011.08.009
http://www.c-a-m.org.cn/CN/Y2011/V1/I8/29
