Publications

Abstract : Due to difficulties including a cluttered background, partial occlusion, and variations on dimensions, angle, illumination, or look, identifying human endeavors using video clips or still images is a challenging process. A numerous mechanism for recognizing activity is necessary for numerous applications, such as robotics, human–computer interaction, and video surveillance for characterizing human behavior. We outline a classification of human endeavor approaches and go through their benefits and drawbacks. In specifically, we classify categorization of human activity approaches into the two broad categories based on whether or not they make use of information from several modalities. This study covered a depth motion map-based approach to human recognizing an action. A motion map in depth created by adding up to the fullest differences with respect to the two following projections maps for each projection view over the course of the entire depth video series. The suggested approach is demonstrated to be computationally effective, enabling real-time operation. Results of the recognition using the dataset for Microsoft Research Action3D show that our method outperforms other methods.