Fall detection scheme based on temperature distribution with IR array sensor

Abstract

In recent years, the rate of older people has been increasing with the aging of population. In such a situation, many older people are injured by falls every year. As a countermeasure, several studies have been proposed to inform others of a fall as a quick post-fall treatment. Therefore, fall detection methods using various sensors have been proposed. Assuming fall detection at places such as home or hospital room, a fall detection method without invasion of privacy and wearing anything is required. In this paper, we propose a fall detection method using IR array sensors. The method allows for fall detection that is inexpensive and capable of privacy protection in a non-wearable form. Also, we analyze temperature distributions using machine learning to enable quicker and more accurate fall detection. We evaluate multiple algorithms of machine learning to select best algorithm. Then, classifiers are created based on these algorithms. We calculate and compare the accuracy of these classifiers. One of the learning data is a series of temperature distribution data for 2 seconds. One temperature distribution is acquired every 0.1 seconds by IR array sensors installed on a ceiling. We prepare 1600 learning data (400 series of learning data each with 4 actions: fall, walking, lying, and none). Based on these data, classifiers are performed using multiple algorithms to determine accuracy. The most accurate algorithm is Voting classifier with 97.75% accuracy. Therefore, the evaluation result showed that the proposed method is possible to classify with high accuracy using IR array sensors based on these prepared learning data.