Human Activity Recognition in Untrimmed Video using Deep Learning for Sports Domain

Abstract

Rapid changing in technology automated human activity recognition most research topic nowadays. As motion camera very problematic for monitoring the human activity in different area like in workplace, office, industry, hospitals and so on in various application. Tracking and take care of the health of the live camera and evaluate the operation done correctly or not. The computer vision area to find the activity done by a human with the help of object detection, feature extraction, cluttered background, occlusion and applying deep learning approaches to reach the solution. In this paper considering the sports activity because the sports activity is the combination of the different sub-activity present. To identify the main activity based on the sub-activity, most of the sports game observed that having the starting activity will be the same and but next sub activity of a sports game will different of differentiated the sports game like high-jump, long-jump, cricket bowling, all this sport game starting with the same sub-activity i.e. running, but second activity differentiate the game, in high-jump is ’high jump after running’, in Long-jump is the ’long jump after running’, and in cricket, bowling is ’bowling after the running’. Video is one of the strongest media for knowledge and transcends the field of sport. How instructors or trainers at any level may use video analysis to focus on any variances in human activity in sport specially in competition and training recording where as well as how to enhance efficiency. Video is a grat source of knowledge, and using it correctly will significantly improve any mentor willing to do the right thing. So, all second sub-activity differentiate the sports game. This is the main objective performance based to find sub-activity and recognized the correct and improve main sport activity using sub-activity using deep learning methodology CNN and LSTM algorithm will discuss in the paper.