Human behavior Analysis, using visual information in a given
image or sequence of images, has been an active area of research
in computer vision community. The image captured by
conventional camera does not provide the suitable information to
perform comprehensive analysis. However, depth sensors have
recently made a new type of data available. Most of the existing
work focuses on body part detection and pose estimation. A
growing research area addresses the recognition of human actions
based on depth images. In this paper, an efficient method for
human action recognition is proposed. Our research makes the
following contributions: the proposed method makes an efficient
representation of human actions by constructing a feature vector
based on the human’s skeletal information extracted from depth
images. Then, introducing these feature vectors to Multi-class
Support Vector Machine (MSVM) to perform the action
classification task. The proposed representation of the human
action ensures it is invariant to the scale of the subjects/objects and
the orientation to the camera, while it maintains the correlation
among different body parts. A number of experiments have been
performed in order to evaluate the proposed algorithm. The results
revealed that the proposed algorithm is efficient and leads to an
improved action recognition process. Moreover, it is suitable for
implementation in a real-time behavior analysis. |
