Joint Orientations from Skeleton Data for Human Activity Recognition

The recognition of activities performed by humans, in a nonintrusive and non-cooperative way, is a very relevant task in the development of Ambient Intelligence applications aimed at improving the quality of life by realizing digital environments that are adaptive, sensitive and reactive to the presence (or absence) of the users and to their behavior. In this paper, we present an activity recognition approach where angle information is used to encode the human body posture, i.e. the relative position of its different parts; such information is extracted from skeleton data (joint orientations), acquired by a well known cost-effective depth sensor (Kinect). The system is evaluated on a well-known dataset CAD- 60 (Cornell Activity Dataset) for comparison with the state of the art; moreover, due to the lack of datasets including skeleton orientations, a new benchmark named OAD (Office Activity Dataset) has been internally acquired and will be released to the scientific community. The tests confirm the efficacy of the proposed model and its feasibility for scenarios of varying complexity.