基于音频分析的视频场景检测  

Inspection of Video Frequency Scene Based on Audio Frequency Analysis

作　　者：倪宁[1] 卢刚[1] 卜佳俊[2] 

机构地区：[1]杭州科技职业技术学院计算机系,浙江杭州310022 [2]浙江大学计算机科学与工程系,浙江杭州310027

出　　处：《计算机仿真》2006年第8期184-187,195,共5页Computer Simulation

摘　　要：目前场景检测的研究,主要是基于图像和视频。但音频同样具有丰富的场景信息,基于音频分析的计算量是比较少的,对自动或者半自动的场景检测,基于音频分析的方法也是更为让用户接受的。可以把基于音频分析的方法作为视频场景检测的辅助手段,以获得更为准确的场景检测和分割。该文提出了一个基于内容的音频分析系统,对视频序列实现基于音频分析的场景检测和分割。该系统能有效的解决许多诸如图像变化了,而实际场景并未变化的情形。且本系统整体运算复杂度较基于视频/图像的场景检测与分割系统要低。The present study on scene detection is mainly based on image and video sequence. Audio sequence, however, also has rich information. Therefore, a method based on audio analysis can be applied as an auxiliary, so as to get more accurate scene detection and split. In this paper, a content - based audio data analysis system is proposed, which accepts video sequence and implements scene detection and segmentation from audio part. The system can solve problems such as changing of image without changing of scene, and the complexity of operation in this system is much lower than that of the system based on image and video sequence analysis.

关 键 词：音频分析 场景检测 音频特征 隐马尔可夫模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]