基于改进迁移学习的篮球运动员动作识别方法

打开文本图片集

文章编号：1003-6180（2026）02-0071-06

[中图分类号]TP391.41;TP18

[文献标志码]A

A Method for Recognizing Basketball Player based on Improved Transfer Learning

WU Linlin 1 ， YANG ZHenxing 2

（1. School of Physical Education， Chizhou University， Chizhou 247000， China; 2. School of Physical Education， Fuyang Normal University， Fuyang 236000， China）

Abstract： Propose basketball player recognition method based on improved transfer learning——Improve the recognition method of transfer learning. Extracting high-dimensional invariant features of athletes' motion trajectories using Gaussian mixture models， and combining K-means clustering to screen the source domain features closest to the target domain; Build a transfer learning model that integrates attention mechanism and hybrid kernel support vector machine， weights key action features， and introduces hybrid kernel support vector machine to improve classification performance. The experimental results show that the improve the recognition method of transfer learning has practical recognition results in basketball action recognition， and can achieve high-precision and low-energy sports training action recognition， providing reliable technical support for sports training monitoring and analysis.

Keywords： improved Transfer learning ; sports training athlete; action recognition ; image enhancement

运动员动作是体育训练的核心，动作规范性决定了训练质量与运动表现。（剩余6274字）