摘要
为克服单个行为表达方法有效性上的不足,提出了一种基于多特征融合和支持向量机(SVM)的人体行为识别(HAR)方法。首先,利用背景差分提取运动显著区域;然后提取运动显著区域的剪影直方图和光流直方图,并采取一定的融合策略,构建融合特征结合SVM识别人体行为。实验以广泛使用的公开数据集Weizmann为研究对象,正确识别率达到99.8%以上。结果表明,提出的特征融合及识别方法能有效地对人体行为进行识别;而且,由于规避了比较耗时的序列匹配操作,减少了计算量。
Human action recognition （HAR） has become and pattern recognition due to a wide range of promising one of the most active topics in computer vision applications. In order to overcome the deficiency of single representation method,a new recognition algorithm of human action based on multi-feature fusion and support vector machine （SVM） is presented in this paper. The proposed algorithm consists of three essential cascade modules. First, the human silhouette is obtained by separating the salient regions and the background based on background subtraction. Then,the multi-feature fusion is constructed by using two types of available features, the histogram of the silhouette and the optic flow. The human activi- ty recognition can commonly be viewed as a multiclass classification problem. Finally, the multiple features are sent to the SVM for recognizing the human activity. The experimental results show that the proposed method can achieve the correct recognition rate above 99.8% for the Weizmann benchmark data set. Inter-related analyses conclude that the proposed algorithm is effective and promising. The recognition performance of the SVM classifiers and some other mainstream classification techniques is also compared,which further verifies the effectiveness of the proposed algorithm.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2014年第9期1818-1823,共6页
Journal of Optoelectronics·Laser
基金
教育部重点科研项目(108174)
教育部博士点基金(20130191110021)资助项目
