Indian Journal of Science and Technology

Abstract

Objective: The aim of smart home is to create an environment that is aware of the activities of elderly, disabled people within home and then predicting their behavior which aids for further actions like alerts. Predictive Intelligence environment gathers information from Wireless Sensor Networks (WSN) from various parts of the home which includes daily activities, interactions with the objects within the monitoring environment. Methods/Statistical Analysis: Assistance independent living of the elderly helps them to lead the daily life independently in a self-regulating way. Based on the key daily living activity like preparing food, showering, walking, sleeping, watching television, reading books etc., their routine and their wellness can be tracked. Behavior of occupant of smart home different times using prediction methods is collected based on which the extraction of patterns is done leading to classification of activity and rating the activity as normal or abnormal. A novel behavior prediction model for daily activity and analysis in monitoring has been designed and developed. Findings: The datasets are being experimented with the support vector machines and deep learning networks. Based on the best performance results, deep learning network with SVM linear kernel is capable of making correct classification of datasets accurately with the average accuracy of 88.20% and the prediction time is 5.178 sec. The results show that daily normal and abnormal patterns can be identified with behavioral changes. Application/Improvements: The deep learning model of smart home system is a useful approach for learning the mobility habits at the home environment, with the potential to detect behavior changes that occur due to health problems. A deep learning based system can successfully identify and predict the activity of the elder people.

Keywords: Predictive Analytics, Senior Analytics Care, Smart Home, Wearable Sensors