Amirkabir University of Technology AUT Journal of Electrical Engineering 2588-2910 53 1 2021 06 01 A Comparison of Two Neural Network Based Methods for Human Activity Recognition 17 26 4148 10.22060/eej.2020.18517.5357 EN Saeedeh Zebhi Yazd University 0000-0002-2281-7316 Seyed Mohammad Taghi AlModaressi Yazd University 0000-0002-9021-4293 Vahid Abootalebi Yazd University Journal Article 2020 05 31 In this paper, two different methods of human activity recognition based on video signals are introduced. The first method explores the effectiveness of combining feature descriptors obtained by local descriptors and artificial neural network classifier. It is used in the traditional approach and the local descriptors extract interest points or local patches from the videos, and the feature vectors are later constructed based on the intrests, and eventually feature vectors are used as the input of a two-layer feed-forward artificial neural network (ANN). Experimental results show that using the HOG3D descriptor with ANN gives the best performance. On the other hand, deep learning architectures have attracted much consideration for automatic feature extraction in the last years, so an improved 3D convolutional neural network architecture is also designed as the second method. They are implemented and compared with state-of-the-art approaches on two data sets. The results exhibit that method 1 is superior when the shortage of sample data is the main restriction. It respectively achieves recognition accuracies of 97.8% and 99.8% for the Weizmann and KTH action data sets. In addition, method 2 is considerable for its automatic features extraction, and achieves an acceptable result with lots of original training data. As a result, it gets recognition accuracy of 92% for the KTH data set while this value is drastically reduced for the Weizmann data set. https://eej.aut.ac.ir/article_4148_810dfbbebb17302018ae903e9cb7a483.pdf