Research Based On The Energy Consumption Estimation Model Of Specific Resistance Training Actions For Multi-source Data Fusion

Research purposes:The estimation of exercise energy consumption has always been a hot issue in the research of kinematics at home and abroad.Most of the existing energy consumption prediction models mainly predict long-term aerobic exercise,and the unit of energy consumption estimation is usually calories/min.However,resistance training uses the number of action repetitions instead of the duration of the action as the cumulative sign of the training volume,which is not suitable for the existing fitting model.Therefore,this study proposes an energy consumption estimation model suitable for resistance training,using polynomial feature engineering and Lasso regression to construct two fitting models of upper limb dominance and lower limb dominance,so as to realize the resistance training exercise energy consumption of young people.estimate.Research method:This study recruited 32 undergraduates and graduate students aged 17-32 years old as the experimental subjects(16 males and 16 males).In the experiment,the subjects wore a gas metabolism analyzer,a heart rate armband and three on their dominant wrists and waist.,The IMU of the dominant instep records the data generated by 24 resistance training actions completed 10 times respectively.After processing,the collected raw data is converted into a data set containing 363 action samples.The data set is divided into a training set and a validation set at a ratio of 8:2.The training set is used to train the resistance training energy consumption estimation model,and the validation set is used to verify the accuracy and consistency of the model.Research result:In this study,four models were obtained through different data source input combinations and algorithm framework combinations:unclassified heart rate fitting model,unclassified heart rate IMU fusion model,upper limb and lower limb heart rate IMU fusion model.After the model is built,the verification set is on the above four models:the root mean square errors are:2.840,2.318,1.447,and 1.567;the goodness of fit are:0.610,0.740,0.897,and 0.877;Pearson’s correlation coefficient(P<0.01)respectively:0.870,0.893,0.957 and 0.937;the residual mean values of the Bland-Altman diagram are:-1.3,-1.1,0.6 and 0,respectively.Research conclusion:(1)In the problem of energy consumption estimation for resistance training,the model that adds the IMU data source as the input variable can improve the accuracy and consistency of the model compared with the traditional heart rate model.(2)In the energy consumption estimation problem of resistance training,the algorithm framework using classification fitting can improve the accuracy and consistency of the model compared with the algorithm framework without classification.(3)The goodness of fit of the upper limb heart rate IMU fusion model and the lower limb heart rate IMU fusion model constructed in this study can reach above 0.85,and the Pearson correlation coefficient can reach above 0.90,which can be used in sports equipment to achieve comparison.Accurately estimate the energy consumption of resistance training.