Human activity recognition with different artificial neural network based classifiers

Human Activity Recognition is a popular topic of research, with the importance it carries and its limited feature vector, to reach high success rates because of the difficulty faced in classification. With the increase of movement measurability for individuals via inertia measuring units embedded inside the smartphones, the data amount increases which lets new classifiers to be designed with higher success in this field. Artificial neural networks can perform better at such classification problems in comparison to conventional classifiers. In this work, various artificial neural networks have been tried to form a classifier for the University of California (UCI) Human Activity Recognition dataset and resulting success rates for those classifiers are compared with existing results for same dataset in the literature.