Machine learning Based on Body Points Estimation for Sports Event Recognition
Naif Al Mudawi, Muhammad Tayyab, Muhammad Waqas Ahmed, Ahmad Jalal
Abstract
To examine the realm of event identification and recognition within sequential images, various parameters have been employed. These parameters encompass factors like the size, location, or positioning of human body parts and their associated contextual influences. we conducted estimations of multiple key points on the body to effectively monitor and track its presence within intricate events. To identify these key body points, we employ a fusion of diverse feature descriptors, combining FAST and Akaze, in a harmonious manner. This fusion enhances the discriminative power of the extracted features, making the identification process more robust and effective. For the classification phase, a hybrid approach is introduced, synergizing Particle Swarm Optimization (PSO) and Artificial Neural Networks (ANN). In our extensive evaluation of the proposed approach, we conducted experiments using the UCF-101 dataset, achieving a noteworthy accuracy of 85.9%. Moreover, our precision score reached 84.1%, while the recall rate was 86.6%.