Subject posture recognition by Support Vector Machine Using Obrid-Sensor

Abstract

Falling in elderly people are a common cause of severe injury. Especially falling among elderly living alone, there is a high risk of serious accidents due to delayed detection of the accident. Thus, a system that can detect falling in a private room is expected. In this study, we propose the method to detect standing and falling of the subject. Our detection system is based on the brightness data of detection space. In the proposed method, Support Vector Machine (SVM) is applied to brightness data obtained from a one-dimensional brightness distribution sensor (Obrid-Sensor) to detect falling. In the previous method detected standing and falling by applying Deep Neural Network (DNN). The proposed method was able to detect standing and falling with a few data and calculating cost than the previous method. As a result, a high distinction rate of the subject's falling and standing state was 97.3 %.