An Approach To Enhance Fall Detection Using Machine Learning Classifier

Abstract

Environmental protection is a concept that uses information and communication technology(ICT) for enhancing people's daily lives. Human fall detection is an effective environmental support sub-area. One of the critical issues in elderly people is the fall detection. In this work, we have proposed the fall detection algorithms using machine learning which later uses a fog computing approach to send information to the caregiver in real-time. One class method based on a support vector machine is used to build fall detection and a Smartphone accelerometer is used for the data collection. We have considered five features from the Smartphone accelerometer for the building of the fall detection model. Fog computing approach is used to intimate the caregiver about the fall in real-time through the cloud connection even in the absence of fog node. The innovation is that we have used the multiplication of the kernel matrix to enforce a one-class classification. During the detection of human fall, the model has achieved 100% sensitivity and 98.8% specificity.