视频检索中基于GMM聚类的无监督情感场景检测  

Unsupervised ESD Method Based on GMM Clustering in Video Retrieval

作　　者：齐兴斌[1] 赵丽[2] 李雪梅[1] 田涛[3] 

机构地区：[1]山西大学计算机工程系,山西太原030013 [2]北京航空航天大学计算机学院,北京100083 [3]北京师范大学教育信息技术协同创新中心,北京100875

出　　处：《电视技术》2015年第5期131-136,共6页Video Engineering

基　　金：国家自然科学基金项目(61202163);山西省自然科学基金项目(2013011017-2);山西省科技攻关项目(20130313015-1)

摘　　要：为了高效地从视频中检索出激动人心的场面,提出了一种基于高斯混合模型的无监督情感场景检测方法。首先,从面部选取42个特征点,并定义10种面部特征;然后,利用高斯混合模型将视频的帧划分为多个聚类;最后,利用每一帧的面部表情分类结果将情感场景划分为单个聚类,并通过场景集成和删除完成检测。在生活记录视频和MMI人脸表情数据库上的实验结果表明,该方法的检测率、分类率分别高达98%,95%,检测5分钟左右的情感场景视频仅需0.138 s,性能优于几种较为先进的检测方法。For the purpose of an efficient retrieval of impressive scenes from videos, an emotional scene detection method based on Gaussian mixture model is proposed. Firstly, 42 feature points are selected from facial, and 10 features are defined. Then, Gaussian mixture model is used to divide video be multiple clusters. Finally, emotion scene is divided into single cluster by using facial expression classification results of each frame, and scene integrating and deleting is used to finish detecting. Experimental results on life record video and MMI face expression database show that the detecting and classification rate of proposed method can achieve 98% and 95% respectively. It takes only 0.12 seconds in detecting emotion scene video with five minutes. Proposed method has better performance than several advanced detecting methods.

关 键 词：视频检索 情感场景检测 面部表情识别 无监督 高斯混合模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]