With advancements in AI, smart home gyms are becoming increasingly popular for providing fitness assistance in indoor environments. In this research, we propose a layer-by-layer framework, called Wi-Fitness, which bridges video perception with Wi-Fi sensing for smart fitness. At the data preprocessing layer, the singular value decomposition-based channel state information denoising mechanism is leveraged to do the Wi-Fi data calibration. Diverse and high-quality training samples are generated by a random quantization-based data augmentation method. At the bimodal fusion layer, the heterogeneity between the Wi-Fi and video is mitigated by the local attention mechanism and the bimodal feature integration mechanism. For the video modality, the attention-based spatio-temporal graph convolutional network (AST-GCN Net) is proposed to refine spatial information. The spatio-temporal semantic alignment module is proposed to transfer spatial information from video to Wi-Fi and maintain temporal consistency across modalities. The fitness assessment layer provides exercise visualization. The generalization of Wi-Fitness is enhanced by layer-by-layer collaboration. Wi-Fitness demonstrates its effectiveness by achieving an average F1-Score of 92.68% in three typical indoor environments.

Background: E-sports gained growing recognition as a competitive pursuit with quantifiable physiological demands. While previous studies have shown that cognitive stress modulates energy metabolism and autonomic nervous system regulation, phase-specific physiological responses during gameplay remain poorly understood. Methods: 20 elite League of Legends players (rank ≥ Platinum) were enrolled. Their cardiopulmonary function and autonomic nervous system activity were monitored during rest and across three distinct match phases (early-, mid-, and late-game). Energy metabolism parameters were measured using a portable cardiopulmonary testing system, and heart rate variability indices were assessed with a chest-worn monitor. Results: Compared with the resting state, the early-game elicited significant increases in energy expenditure oxygen consumption (VO ₂ ), carbon dioxide production (VCO ₂ ), and respiratory exchange ratio (all p < 0.001), with carbohydrate oxidation accounting for approximately 63% of total energy supply. Heart rate (HR) increased by 12.8%, while the root mean square of successive differences (RMSSD) rose by 52.2%, indicating sympathetic-parasympathetic coactivation. In the mid-game, metabolic indices declined but remained above baseline levels, characterized by sustained carbohydrate dominance (about 63.6%) and increased fat oxidation (about 30.2%); heart rate variability indices reflected sympathetic predominance accompanied by partial parasympathetic recovery. The late-game was characterized by slight rebounds in metabolic load and carbohydrate utilization (about 68.2%), accompanied by decreased Heart rate and elevated RMSSD, suggesting partial autonomic recovery alongside incipient neural fatigue. Conclusion: Elite e-sports athletes demonstrate dynamic, phase-dependent alterations in energy metabolism and autonomic nervous system regulation. The early phase is characterized by carbohydrate-dominated physiological activation, the mid phase by metabolic stabilization amid sustained cognitive demand, and the late phase by partial autonomic recovery with cumulative neural fatigue. These findings highlight the physiological mechanisms underlying E-sports performance and provide actionable insights for optimizing training regimens, fatigue monitoring protocols, and recovery interventions.

To explore the correlation between the characteristics of lower limb muscle synergies and landing loads during the single-leg landing process in humans. Lower limb electromyographic data were collected from 21 male athletes during single-leg landing. The non-negative matrix factorization algorithm was used to extract analytical indicators, including the number of participating muscle synergies, the time to peak activation of muscle synergies, and the relative muscle weights. Additionally, the associations between the above indicators and the loads during human single-leg landing were analyzed. In the number of synergies, three types of muscle synergies were identified during single-leg landing. The number of synergies showed a significant negative correlation with peak landing impact force and lower limb stiffness(P < 0.05). In the time to peak activation, the synergy 1 of peak activation time was significantly positively correlated with lower limb stiffness (P < 0.05). The synergy 3 of peak activation time was significantly positively correlated with peak impact force (P < 0.05). In the muscle relative weights, the synergy 1 of the relative weight of the rectus femoris was significantly negatively correlated with peak impact force (P < 0.05), while the semitendinosus was negatively correlated with maximum loading rate (P < 0.05). The Synergy 3 of the relative weight of the tibialis anterior was significantly positively correlated with peak impact force (P < 0.05). (1) Increasing the number of participating synergies can stimulate more muscle activity, which contributes to finer coordination and distribution of the impact forces on the body during single-leg landing. (2) Faster activation of synergies governing the landing absorption phase enhances eccentric control of knee extensors, reducing lower limb stiffness. Conversely, quicker activation of synergies dominant in the initial landing impact phase improves calf muscle engagement for energy dissipation. (3) Enhancing the eccentric strength of the rectus femoris during the landing absorption phase reduces muscle stiffness, promoting knee flexion buffering. However, excessive tibialis anterior activation should be avoided to prevent premature compensatory dorsiflexion at the ankle joint.

1 In recent years, major international esports events, such as the esports competition at the Hangzhou Asian Games and the Esports World Cup in Riyadh, Saudi Arabia, have been held worldwide. This ...