| As one of the most popular research topics in action recognition field,fall detection has attracted a lot of interests.This paper focuses on extracting representations with strong description ability from a RGB-D image sequence as well as judging whether a fall happened.Towards this goal,a novel key point trajectory model is proposed for fall detection by representing a fall action as a series of trajectory descriptors.In this proposed model,16 key points including 14 skeleton points and 2 centers of body parts are extracted from each pair of RGB and depth images.Then a global trajectory descriptor is constructed on 16 trajectories that are obtained by connecting the key points across several frames in the RGB-D sequence.The trajectory descriptor incorporates the spatial,depth,and temporal context of key points and characterizes the global motion of human over a short period of time.A random forest is employed to learn the classifier of trajectory descriptors,and an integration rule is developed for detecting falls according to the classification results of all trajectory descriptors within a video.We present an unsupervised triangle fitting model based on 3D skeleton points,which only utilizes five skeleton points and fits them in 3D space.It is observed that the fitting triangles of falls are obviously different from those of other actions.Accordingly,fall detection is achieved in terms of two angle features extracted from the fitting triangle.The proposed method doesn't require training videos with annotations and thus has good generalized performance.In order to evaluate the performance of the proposed methods,experiments were conducted on the UR Fall detection dataset and the SDU fall detection dataset.The unsupervised triangle fitting model based on 3D skeleton points achieves comparable performance with state-of-the-art methods,and the key point trajectory model outperforms the comparison methods.Experimental results demonstrate the effectiveness of the proposed methods. |
PDF Full Text Request