Feature Based Human Activity Recognition using Neural Network

Abstract

In various research areas, human activity recognition system (HAR) becomes more popular. Features is one of the important things to recognize the object or activity in machine learning. Increasing the old age population in our country, the smart nursing home monitoring and heath care supporting system is still needed to develop. The dataset for experiment is collected from 'Cherry Myay' nursing home in Pyin Oo Lwin, Myanmar. We proposed new feature extraction method, new morphological operation method 'Vertical Binary Bridge' pose classification method using SVM and new frequency based feature selection method. Artificial neural network is used to recognize the human activity and abnormal behaviour detection system. Compared with other exiting methods, our proposed system can achieve acceptable result.