Human action recognition using Spatio-temporal Histogram of Structure Tensors descriptor

In recent years, human action recognition has become an important area of computer vision research. Its goal is to automatically analyze ongoing activities from a video or a sequence of image frames in order to correctly classify the videos and corresponding related activity category. In general cases of human action recognition, video is segmented to contain only one execution of a human activity. In the second step, the continuous recognition of human activities must be performed by detecting the starting and ending times of all occurring activities from an input video. In fact, the ability to recognize complex human activities from videos enables the construction of several important applications. Automated surveillance systems in public places, such as airports and subway stations, require the detection of abnormal and suspicious activities, as opposed to normal activities. The recognition of human activities also enables the real-time monitoring of patients, children and elderly persons.In this paper, we introduce a new Spatio-Temporal feature descriptor to extract the local information of a frame. Our work is based on the structure tensor using an orientation tensor which represents a local orientation of a real symmetric metrics. We also propose a new system for human action recognition based on Spatio-Temporal Histogram of Structure Tensor descriptor based on experimental results and we demonstrate that our system overcomes the existing methods in the literature in terms of precision and accuracy.