Decision Trees for Human Activity Recognition Modelling in Smart House Environments

Abstract

Human activity recognition in smart house environments is the task of automatic recognition of physical activities of a person to build a safe environment for older adults or any person in their daily life. The aim of this work is to develop a model that can recognize abnormal activities for assisting people living alone in a smart house environment. The idea is based on the assumption that people tend to follow a specific pattern of activities in their daily life. An open source database is used to train the decision trees classifier algorithm. Training and testing of the algorithm is performed using MATLAB. The results show an accuracy rate of 88.02% in the activity detection task.

Viola: Detecting Violations of Behaviors from Streams of Sensor Data

“…The approach takes as input a set of annotated traces used to train a conditional random field, which then operates on an unlabeled log to abstract activities. Moving to the activity recognition field, Sanchez et al [12] propose a technique to recognize abnormal activities. They use a labeled event log to train a classifier, which is then tested on an unlabeled log.…”