Using wearable technology to optimize sports performance and prevent injuries

Purpose: Wearable devices, as emerging computing platforms, have gradually penetrated into people’s daily life, especially in the field of medical health management showing excellent potential. Methods Motion state recognition is performed by deep fusion CNN-LSTM model, CNN is used to obtain the most representative feature information characteristics of the local space of the motion data, while the LSTM layer is used to capture the long-term temporal correlation of these local features, and both of them are combined to obtain the more representative temporal-spatial correlation transportation state feature information implicit in the wearable gait data. An injury prevention method for exercise example parameters is designed, including patient training load characterization, and a Bi-LSTM network structure is used to design lightweight acceleration features to predict abnormalities in exercise physiological indicators. Results: Monitoring parameters such as heart rate rise slope, 1-minute heart rate recovery value, blood oxygen drop area, and 1-minute oxygen saturation recovery value, the false alarm rate of wearable device health data warning were kept at 2.55%. After exercise status and detected abnormalities in physiological parameters, personalized breathing training was performed, and the contribution ratio of abdominal breathing increased by 27% after training, and the patient’s heart rate decreased by 8.5 bpm and oxygen saturation increased by 2.4% compared to the pre-training period. Conclusion: The methodology in this paper can be more comprehensively optimized for sports performance and injury prevention, and is widely applicable in practical applications.