Title: Health status recognition method of humans in sports videos based on deep learning

Authors: Xunxing Liu

Addresses: College Information and Finance, XuanCheng Vocational and Technical College, Xunancheng, 242000, China

Abstract: When using traditional recognition methods, the problem of low confidence, high complexity, and large mean error increases with the increase of recognition times. A deep learning based method for recognising the health status of people in motion videos is proposed. First, determine the global spatial information of frame difference features and use two-dimensional convolution to extract frame difference features of human health status. Second, based on the frame difference features and combined with normalisation methods, different temporal and spatial distances are converted to determine the facial expression information feature frame differences of the character's health status, completing the segmentation of the character's health status frame differences in sports videos. Then, by using the segmentation results as model inputs to reduce the fitting degree of the recognition results, utilising the fully connected layer in deep learning to output the recognition results, and introducing a loss function to optimise the recognition results, state recognition is achieved. Finally, conduct experimental verification. The experimental results show that the proposed method has high confidence, low complexity, and small error, indicating good recognition performance.

Keywords: deep learning; sports videos; human's health status; identification; frame difference segmentation.

DOI: 10.1504/IJBM.2026.151093

International Journal of Biometrics, 2026 Vol.18 No.1/2/3, pp.128 - 147

Received: 27 Nov 2024
Accepted: 11 Apr 2025
Published online: 13 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article