Digital Twin technology has been rising in popularity thanks to the popularity of machine learning in the last decade. As the life expectancy of people around the world is increasing, so is the focus on physical activity to remain healthy especially in the current times where people are staying sedentary while in quarantine. This article aims to provide a survey on the field of Digital Twin technology focusing on machine learning and coaching techniques as they have not been explored yet. We also define what Digital Twin Coaching is and categorize the work done so far in terms of the objective of the physical activity. We also list common Digital Twin Coaching characteristics found in the articles reviewed in terms of concepts such as interactivity, privacy and security and also detail future perspectives in multimodal interaction and standardization, to name a few, that can guide researchers if they choose to work in this field. Finally, we provide a diagram for the Digital Twin Ecosystem showing the interaction between relevant entities and the information flow as well as an idealization of an ideal Digital Twin Ecosystem for team sports’ athlete tracking.

Exercise is a prevailing topic in modern society as more people are pursuing a healthy lifestyle. Physical activities provide significant benefits to human well-being from the inside out. Human pose estimation, action recognition and repetitive counting fields developed rapidly in the past several years. However, few works combined them together to assist people in exercise. In this paper, we propose a multitask system covering the three domains. Different from existing methods, heatmaps, which are the byproducts of 2D human pose estimation models, are adopted for exercise recognition and counting. Recent heatmap processing methods have been proven effective in extracting dynamic body pose information. Inspired by this, we propose a deep-learning multitask model of exercise recognition and repetition counting. To the best of our knowledge, this approach is attempted for the first time. To meet the needs of the multitask model, we create a new dataset Rep-Penn with action, counting and speed labels. Our multitask system can estimate human pose, identify physical activities and count repeated motions. We achieved 95.69% accuracy in exercise recognition on the Rep-Penn dataset. The multitask model also performed well in repetitive counting with 0.004 Mean Average Error (MAE) and 0.997 Off-By-One (OBO) accuracy on the Rep-Penn dataset. Compared with existing frameworks, our method obtained state-of-the-art results.