摘要
流媒体中基于语音的实时语义检测是信息安全的重要方面,在分析研究基于音频的内容检索技术基础上,提出了一种适合于流媒体中基于语音的实时语义检测方法,该方法通过预判决和二次匹配两个过程,利用特征值前向序列对视频的语音特征进行加权匹配。实验结果表明,利用该匹配算法能满足语音实时语义检测的要求,可以更好地获得流媒体检索所需要的实时性。
Real-time audio semantic detecting in stream media is an important part of information security. In this paper, an improved algorithm of features matching is introduced by analyzing the technique of content retrieval based on audio to show a method that is fit for real-time audio semantic detecting in stream media. Thisalgorithm includes two steps： pre-adjudgement and re-matching, it uses forward sequence of features to implement weighted matching. The experimental results show that speed of content retrieval is faster, so that more real-time capability which is important in stream media retrieval is obtained.
出处
《信息技术》
2009年第3期54-57,共4页
Information Technology
关键词
语音检索
流媒体
信息安全
特征匹配
FDDM
audio retrieval
stream media
information security
feature matching
FDDM
