Multi-Resident Activity Monitoring in Smart Homes: A Case Study

Abstract

This paper demonstrates a system that can turn a normal house to a smart house for daily activity monitoring with the use of ambient sensors. We show a proof of concept which includes a method of annotation using voice recording and deep learning technique for automatic recognition. Our multi-resident activity recognition system (MRAR) is designed to support multiple occupants in a house with minimum impact on their living styles. We evaluate the system in a real house lived by a family of three. The experimental results show that it is promising to develop a smart home system for multiple residents which is portable and easy to deploy.