Development of a Healthcare Application on Smartphones Utilizing Human Activity Recognition

In recent years, with the growing health consciousness, the demand for high-calorie-burning exercises, particularly High-Intensity Interval Training (HIIT), has been increasing. This study aims to improve the quality of HIIT by developing a healthcare application on smartphones that addresses the following two points: (1) There is no guarantee that the exercises are performed accurately according to the HIIT instructions. (2) There is no guarantee that the exercise time is measured accurately. In this study, we utilize a Human Activity Recognition (HAR) method based on data obtained from a smartphone's accelerometer (Acc), gyroscope (Gyr), and geomagnetic sensor (Mag). However, HAR using smartphones faces the issue of decreased recognition accuracy due to missing these sensor data caused by power shortages. Therefore, in this study, we develop an HAR method robust against sensor data loss. As a result, we demonstrate that stable recognition accuracy can be maintained even when data from one sensor is missing. Furthermore, by using the proposed application, we verify that (1–2) can be effectively addressed.