基于pLSA的人体动作识别
首发时间:2013-03-11
摘要:提出一种基于主题模型的人体动作识别方法。该方法首先提取时空兴趣点(STIP,space-time interest point)来描述人体运动,然后提出一种时间-梯度直方图(简称T-HOG)算法,该算法对空间梯度直方图(HOG,Histogram of oriented Gradients)仅能描述STIP 3D区域立方体空间信息的不足进行了弥补,实现对STIP 3D区域立方体时间方向变化信息的量化表示,最后使用概率潜在语义分析 (pLSA, probabilistic Latent Semantic Analysis) 模型识别人体动作。同时,针对pLSA隐性主题正确性无法保证的缺点,算法将主题与动作标签"一对一"相关,通过监督方式得到主题,保证了训练中主题的正确性。该算法在KTH人体运动数据库和Weizmann人体动作数据库进行了训练与测试,动作识别结果正确率在91.50%以上。
关键词: 主题模型 动作识别 时空兴趣点 梯度直方图 隐形主题
For information in English, please click here
Human action recognition based on pLSA
Abstract:IA human action recognition method based on a probabilistic topic model is proposed. Firstly, the method extracts space-time interest points to describe human motion, then presents the histogram of oriented gradients in time direction (shorted for T-HOG) to quantify the STIP surrounding 3D volume patch, which makes up the shortage of the spatial histogram of oriented gradients(HOG) that only reflects the spatial information. Lastly human actions are recognized by probabilistic latent semantic analysis (pLSA). For solving the problem of latent topics that are not guaranteed in pLSA, the topics obtained in supervised fashion correspond to action labels one by one. Action recognition results were presented on KTH human motion data set and Weizmann human action data set. Our results show that the recognition rate of the algorithm is more than 91.50%.
Keywords: topic model action recognition space-time interest points HOG latent topic
论文图表:
引用
No.****
同行评议
勘误表
基于pLSA的人体动作识别
评论
全部评论0/1000