User Estimation with Touch Panel Buttons Toward In-Home Activity Recognition

Smart homes that improve people's lives are attention. To realize a smart home, it is necessary to recognize user activity. Existing studies have collected user activity data by installing many sensors in the home, however, the collected activity data does not include user names, and it is not possible to recognize the activity of multiple users while distinguishing between them. In this study, we propose an in-home activity recognition method that identifies users by performing user estimation using data collected from home appliance operations. In this paper, we propose a user estimation method using button operations among various home appliance operations. We evaluated the user estimation performance of the proposed method using button operation data collected from 8 subjects and found that accuracy was 86.4% or higher when the amount of training data was 10 trials and only features obtained from the pressure were used. This result is comparable to existing studies that require multiple sensors, indicating the feasibility of user estimation with a single sensor.