Real-time Dynamic Data Analysis Model Based on Wearable Smartband

Since the traditional annual physical fitness test for adolescents in schools lacks of real-time dynamic data collection and deep analysis, we propose a data analysis model mapping from raw dynamic data to physical fitness evaluation, based on a self-developed wearable smartband by which we collect dynamic data from middle school students in Beijing. Firstly, the model presents a preprocessing algorithm which consists of a smoothness priors approach (SPA) and a median filter (MF), aiming for preprocessing of both photo plethysmo graphy(PPG) signals and three-axis acceleration data collected from the wearable smartband. Secondly, the model implements physiological and physical index estimation to acquire heart rate (HR), blood oxygen saturation (SpO ₂ ) and exercise amount estimates from the preprocessed data. Thirdly, the model extracts several key features closely related to physical fitness evaluation from the estimated HR, SpO ₂ and acceleration data. Finally, a support vector machine (SVM) algorithm is employed for classification of physical fitness level of different smartband users. An application testing of our self-developed wearable smartband has been implemented in Tongzhou No. 6 middle school of Beijing. Experimental evaluation results demonstrate feasibility and effectiveness of both the smartband hardware/software and the proposed data analysis model.