A Smart Automated Complex Motion Recognition Technique

This paper presents a noble appearance-based representation and recognition approach to the recognition of complex motions and gestures of several peoples ’ several actions from uncalibrated multiple cameras and a frontal-view camera. This paper has proposed to extend the basic motion history image (MHI) by evaluating the directionality of motions in an image. It proposes four-directional history image and energy image to evaluate the feature vectors for recognizing several complex actions. In this method, the optical flow is calculated instead of the frame subtraction to calculate the motion history and motion energy templates. It segregates the computed optical flow into four different components according to four directions, namely – top, bottom, left and right directions. From these four optical flow components, four directional motion history images and the corresponding motion energy images have been calculated. One of the great concerns for basic MHI method is that old motion representation may be overwritten by newer motion information for complex action. Therefore, for basic MHI method, the recognition rate is poor for complex motion. The proposed approach can solve this problem significantly by improving the recognition rate compare to the basic method. To calculate the feature vector, this method employs Hu moments and calculates a 64-dimensional feature vector for each motion sequence