Activity recognition of interacting people

In recent years, human activity recognition is becoming more popular in many areas such as human-robot interaction because of easy availability and widespread use of RGB-D sensors. The aim of this study is to automatically recognize human activities with deep learning techniques using three-dimensional skeletal joint data from the RGB-D sensor. Our methods uses the joint data directly and automatically acquires the features to be used in the classification, thus provides superiority to the methods which uses hand-crafted features. In our work, the NTU RGB + D dataset which is quite new and challenging compared to the datasets in the literature, is used. With using 2D, 3D Convolutional Neural Networks and LSTM Networks a performance analysis was performed. As a result of the experiments made, the technique applied by the 3D Convolutional Neural Network achieves the high classification accuracy with by obtaining much more meaningful features compare to the LSTM Network.