Physical Activity Recognition Using Streaming Data from Wrist-worn Sensors
Katika Kongsil, Jakkarin Suksawatchon, Ureerat Suksawatchon · 2019
Most of the existing researches in smartwatches based activity recognition focused on developing the subject (user) specific approach or personal model which the subject must collect the labeled data for training the model. It is inconvenient for the users are unable to perform all activities during the specified times. In this paper, we introduce a cross subjects approach or impersonal activity recognition model based on the fusion of two sensors embedded on smartwatches called S-PAR. It stands for Smartwatches based Physical Activity Recognition. Therefore, the users who utilize the model, are not necessary to gather initial the labeled data. The experiments were carried out to examine the performance of S-PAR model against with state-of-the-art methods by using two public databases collected under realistic conditions. From the results, S-PAR model provides the overall performance in detection and prediction activities type. Therefore, our proposed model can be used in the real life environment.