A Hybrid Approach for Sports Activity Recognition Using Key Body Descriptors and Hybrid Deep Learning Classifier.

Area of Science:

• Computer Vision
• Machine Learning
• Pattern Recognition

Background:

• Event recognition in sequential images is crucial for applications like surveillance and human-computer interaction.
• Accurate tracking and analysis of human body parts and their motion are essential for understanding complex scenarios.

Purpose of the Study:

• To develop an effective approach for event recognition in sequential images.
• To enhance the accuracy and robustness of human action recognition systems.

Main Methods:

• Utilized key body points for tracking and monitoring human presence.
• Applied diverse feature descriptors (MSER, SURF, HOG, Optical Flow) to skeleton points and silhouettes.
• Employed feature fusion and a hybrid Convolutional Neural Network (CNN) + Recurrent Neural Network (RNN) classifier.
• Incorporated Grey Wolf Optimization (GWO) for optimal feature selection.

Main Results:

• Achieved high accuracy rates of 98.5% on the UCF-101 dataset and 99.2% on the YouTube dataset.
• Demonstrated superior performance compared to state-of-the-art methods in event recognition.
• Validated the effectiveness of feature fusion and the hybrid CNN+RNN classifier.

Conclusions:

• The proposed approach effectively recognizes events in sequential images by integrating body part features and context.
• The combination of advanced feature extraction, fusion, and a hybrid deep learning model significantly improves action recognition accuracy.
• This method offers a promising solution for real-world event recognition tasks.