Research on motion recognition and energy consumption prediction of competitive aerobics based on machine learning algorithm and accelerometer

This study combines machine learning technology with accelerometer to explore the motion recognition and energy consumption prediction methods in competitive aerobics single events, aiming to measure the energy consumption characteristics of athletes in competition routines through experiments, and provide theoretical support for physical training, motion design and athlete physical management. The study took a group of competitive aerobics athletes as the subjects, collected motion data using accelerometers and gas metabolism analyzers, screened out the main factors related to energy consumption through correlation analysis and regression modeling, and constructed a prediction model for energy consumption. The results showed that the heart rate, metabolic equivalent and energy consumption per unit time of competitive aerobics single events were at a high level, carbohydrates were the main source of energy, and there were significant differences in physiological indicators such as heart rate, energy consumption and oxygen uptake between athletes of different genders. Further analysis found that there was a strong correlation between the accelerometer data of the right waist and energy consumption, indicating that the data of this part had a high accuracy in predicting energy consumption. The prediction model was tested by a variety of verification methods. The results showed that the model has good prediction accuracy and stability, and can provide effective technical support for physical training and personalized physical management of athletes in competitive aerobics. In general, this study provides a new theoretical tool for physical optimization of competitive aerobics, and also provides a reliable prediction model for the scientific evaluation of athletes' energy consumption, which has high application value and promotion potential.