New approach for action recognition using motion based features

Analyzing the actions of humans by using cameras can be termed as Action Recognition. This concept of Action Recognition is now used in many fields, especially in the field of Robotics and intelligent systems in which there is a greater need for the recognition of the actions. Recognizing of humans and also their activity is very much important for any intelligent system, which is to be done intelligently and effortlessly with a human-inhabited environment. Although many efficient applications are available for the purpose of action recognition, the most active application domains in computer vision which is currently most used is "looking at people". This paper mostly discusses about the usage of Computer Vision and Machine Learning techniques with the help of the toolboxes. Also video-based view-invariant action recognition is proposed as a new discriminative model in this paper. A better combination of invariance and distinctiveness is obtained by perfectly fusing the motion pattern and view invariants together in the proposed discriminative model. A series of issues, including interest point detection in image sequence, motion feature extraction and description, and view-invariant calculation is discussed. Initially an efficient motion sensor method, which is better than traditional background modeling and tracking based methods, is employed for extracting the motion information from videos. Exaction of variety of statistical information from motion and view-invariant feature based on cross ratio is done in the next step of feature representation. By applying a discriminative probabilistic model-hidden conditional random field to model motion patterns and view invariants, we could fuse the statistics of motion and projective invariability of cross ratio in one framework. This step if the final step termed as action modeling.