Related Experiment Video
Updated: Nov 9, 2025
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
9.6K
Applying TS-DBN model into sports behavior recognition with deep learning approach.
Yingqing Guo1, Xin Wang1,2
1Institute of Physical Education, Shandong University, Jinan, China.
The Journal of Supercomputing
|April 12, 2021
Summary
This study introduces a novel deep learning model for human sports behavior recognition, achieving over 90% accuracy. The model outperforms existing methods like Convolutional Neural Networks and Deep Belief Networks on benchmark datasets.
Area of Science:
- Computer Science
- Artificial Intelligence
- Machine Learning
Background:
- Automatic recognition of human sports behavior from video data is crucial for analysis.
- Existing deep learning (DL) models require improvement in handling spatiotemporal features for accurate recognition.
Purpose of the Study:
- To develop and evaluate a novel human sports behavior recognition model using deep learning.
- To enhance spatiotemporal Deep Belief Networks (DBNs) with multi-scale data analysis and pooling strategies.
Main Methods:
- A human sports behavior recognition model was proposed, focusing on spatiotemporal features.
- The model was trained using video frame data from KTH and UCF datasets.
- Performance was evaluated against DBN, Convolutional Neural Network (CNN), and DBN-Hidden Markov Model (HMM) algorithms using TensorFlow.
Main Results:
- The proposed model achieved the highest accuracy, reaching approximately 90%.
- Performance varied across datasets, with KTH generally showing lower accuracy than UCF.
- Specific actions like boxing (KTH) and lifting (UCF) showed high recognition rates, while running (KTH) and walking (UCF) were lower.
Conclusions:
- The proposed model demonstrates superior accuracy compared to classic DL algorithms for human sports behavior recognition.
- The findings provide an experimental foundation for future research in this domain.
