摘要
针对目前体育视频分析缺乏有效的多模式融合分析方法以及没有一个统一的体育视频内容分析框架等问题,提出了3种多模式融合的分析框架,可用于多媒体视频内容分析的统计模型,即FHHMM、CHHMM、PHHMM。研究了贝叶斯动态网络的原理,将事件的关系用该网络的拓扑结构来表示,然后在统计理论和所推导算法的基础上,将多媒体视频中相关的两个事件以概率的方式有机地结合起来,建立多模式交互关系,从而在分析的过程中提高视频分析的有效性。实验证明,PHHMM模型性能最佳,性能比传统模型有很大提高。
In view of the lack of effective multi-mode fusion analysis methods for sports video analysis and the absence of a unified framework for analyzing sports video content,three statistical models for multi-mode video content analysis are proposed,namely,FHHMM,CHHMM,PHHMM.The principle of Bayesian dynamic network is studied.The relationship between the events is represented by the topology of the network.Based on the statistical theory and the deduced algorithm,the two related events in multimedia video are organically combined,multimodal interactions are established to improve the effectiveness of video analytics in the analysis.Finally,experiments show that,PHHMM model performance is the best.performance of the traditional model has been greatly improved.
作者
石庆福
SHI Qingfu(School of Physical Education,Zhengzhou University of Light Industry,Zhengzhou 450002,China)
出处
《实验室研究与探索》
CAS
北大核心
2019年第12期177-181,共5页
Research and Exploration In Laboratory
基金
国家自然科学基金项目(61401404)
关键词
体育视频分析
多模式融合
贝叶斯动态网络
乘积层次隐马尔科夫模型
sports videos analysis
multi-model fusion
Bayesian dynamic network
product hierarchical hidden Markov model(PHHMM)