1. Psychological characteristics and data collection and analysis in boxing
2. Analysis of boxing action classification and recognition model based on CNN assisted by sports psychology

In the world of boxing, the mental game is just as crucial as the physical aspect of the sport. Psychological characteristics play a significant role in determining a boxer’s performance in the ring. Factors such as anxiety level, self-confidence, team identity, and opponent’s attitude can greatly impact a boxer’s success. Understanding and quantifying these psychological traits are essential for personalized training and improving athletes’ competitive level.

To delve deeper into the psychological state of boxers, a study was conducted to design a psychological scale covering four dimensions: anxiety level, self-confidence, team identity, and opponent’s attitude. The study involved distributing questionnaires to 452 boxers, with a recovery rate of 95.35%. The data collected was analyzed using the Likert five-level scale method to gain insights into the psychological characteristics of the participants.

Furthermore, an innovative boxing action classification and recognition model was proposed, combining sports psychology with advanced deep learning technology. The model integrated the BERT algorithm for processing text data related to athletes’ psychological state and the 3D-RESNET network for extracting spatial-temporal features from video images of boxing movements. By fusing text and video data, the model aimed to provide a comprehensive understanding of athletes’ psychology and action state.

The model’s architecture included the use of a self-defined attention mechanism layer to enhance the integration of feature representations from BERT and 3D-RESNET. This attention mechanism allowed the model to capture key information related to boxing actions more effectively, improving its performance and generalization ability.

Overall, the fusion of sports psychology with deep learning technology in the context of boxing action classification and athlete state analysis opens up new research avenues in sports science and human-computer interaction. The innovative approach presented in this study has the potential to revolutionize how boxing movements are classified and understood, paving the way for advancements in athlete training and performance evaluation.