Analysis of Basketball Technical Movements Based on Human-Computer Interaction with Deep Learning.

Area of Science:

• Computer Science
• Sports Science
• Artificial Intelligence

Background:

• Advancements in computer technology drive evolution in sports data analysis.
• Image-based motion recognition offers feasibility for video action recognition.
• Basketball technical action videos present unique characteristics like fixed athletes and homogeneous scenes, offering advantages for analysis.

Purpose of the Study:

• To address the challenges of numerous and complex basketball techniques in action recognition.
• To propose an effective framework for analyzing basketball technical actions from video data.

Main Methods:

• Development of a 3D convolutional neural network (CNN) framework.
• Utilizing two different resolution image inputs for the basketball technical action dataset.
• Implementing advanced algorithmic processes for action recognition.

Main Results:

• The proposed 3D CNN framework demonstrates effectiveness in recognizing basketball technical actions.
• Experimental results validate the algorithmic process's capability on a basketball technical action dataset.
• The approach successfully handles the complexity and variety of basketball techniques.

Conclusions:

• The developed 3D CNN framework is effective for basketball technical action recognition.
• The study highlights the potential of deep learning in analyzing complex sports actions.
• This research contributes to the advancement of automated sports performance analysis.