Recognition of Basketball Collective Behaviors Based on Neural Network Embedding Learning
WANG Zhenhua
Abstract：Defining and identifying different types of activities is an important task in sports intelligence analysis,which can provide better game strategies for players and coaching teams.However,complex collective sports activities often pose challenges in modeling global relationships and understand behavior comprehensively.This paper proposes a deep learning method for identifying collective in team sports by tracking player and ball position information in basketball games.To effectively model player relationships in team sports,a transformer structure is combined with Long short term memory feature embedding,and team-level pooling layers are utilized to activity recognition.Additionally,the scarcity of manual annotations during the training stage is addressed by generating weak labels.The proposed method is evaluated on a tracking dataset of 632 NBA games.Results demonstrate high accurately in recognizing different types of collective activities
Keywords：group activity recognitiontransformerlong short-term memory embedding
Publication Date：2023-11-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 123-128 )