Human Daily Activity Recognition Using Neural Networks and Ontology-Based Activity Representation

Abstract

In real-life people live together in the same place, recognize their activities is challenging than activities of one single resident, but essential to collect information about real life activities inside home, then ease the assisted living in the real environment. This paper presents a multilayer perceptron model and a supervised learning technique called backpropagation to train a neural network in order to recognize multi-users activities inside smart home, and select useful features according to minimum redundancy maximum relevance. The results show that different feature datasets and different number of neurons of hidden layer of neural network yield different activity recognition accuracy. The selection of suitable feature datasets increases the activity recognition accuracy and reduces the time of execution. Our experimental results show that we achieve an accuracy of 99% with the winner method and 96% with the threshold method, respectively, for recognizing multi-user activities.