Sensor-Based Activity Recognition Inside Smart Building Energy and Comfort Management Systems

Abstract

The challenge of Smart Building Energy and Comfort Management (BECM) systems is to schedule home appliances according to users’ comfort requirements, while contributing to an efficient and sustainable use of the available energy sources, supplied by either the electricity grid or Renewable Energy Sources (RES). To this aim, BECM systems have to monitor users’ habits and learn their preferences, so that their actions can be predicted and appliances can be scheduled accordingly.This paper stems from the observation that actions are usually performed in sequences that repeat according to a pattern. Therefore, activities can be recognized and predicted as soon as a pattern of actions is detected. The framework proposed in this paper aims to predict activities by analyzing the sequences of actions detected by sensors deployed in a Smart Building. Furthermore, a correlation between subsequent activities is found so that sequences of activities can be predicted. Simulation results show that activities can be predicted with an accuracy of 74.78%.