Deep learning has become a key technology in enhancing sports performance, offering new possibilities for training optimization. This study aims to (1) explore the research progress on deep learning applications in enhancing sports performance and (2) analyze future trends in this field. Method: Data from 545 English-language publications (2015-2025) were extracted from the Web of Science Core Collection using keywords related to DL (e.g., CNN, LSTM) and sports performance (e.g., athletic training, motion analysis). CiteSpace 6.4R was employed for bibliometric, co-word, and burst-term analyses. 1) Since 2015, research in this field has steadily increased before plateauing, with the most significant rise from 2021 (73 articles) to 2022 (90 articles) and a peak in 2024 (107 articles). Leading contributors include China (206 publications), the U.S., Germany, Japan, Korea, India, and the U.K. 2) ImageNet Classification with Deep Convolutional Neural Networks is the most cited work, introducing AlexNet and demonstrating its performance on large-scale image classification, significantly influencing computer vision. This study explores deep learning’s application in sports performance, with most literature focusing on specific sports. 3) High-frequency keywords in sports include motion analysis, performance, task analysis, kinematics, gait analysis, and sports training, while in computing, they include deep learning, convolutional neural networks, activity recognition, human pose estimation, classification, and wearable sensors. 4) Research trends are reflected in emergent keywords. From 2015–2019, studies emphasized motion detection, image recognition, kinematics, and image motion analysis for improved motion monitoring and assessment. From 2020–2025, a shift toward velocity, task analysis, gait analysis, and three-dimensional displays highlights a focus on individualized assessment and optimized training. Pose estimation, sports training, IoT, and virtual reality remain active, underscoring the role of deep learning in intelligent sports performance analysis. 1) Over the past decade, research on deep learning for sports performance has grown rapidly before stabilizing, with China, the U.S., Germany, Japan, Korea, India, and the U.K. leading the field. 2) The focus has shifted from general motion detection to refined applications like gait analysis, velocity estimation, and task analysis, though most studies prioritize computational methods over sport-specific needs. 3) Research trends emphasize integrating deep learning, including CNNs and pose estimation, into performance analysis. While progress has been made in motion monitoring, further exploration of sport-specific applications is needed. 4) Advancements in AI and sports science will drive intelligent training, injury prevention, and performance optimization, benefiting both competitive sports and public health.
Yang et al. (Thu,) studied this question.