A deep learning model to recognize human movement by using auto coders

Abstract

In the recent years, there has been great interest in identifying human movement using deep learning and skeletal data for its effective performance compared to models that depend on images or depth data. In this research, we presented a model that solves the problem of the size and inconsistency of data that used on deep learning model and enables these results to works in environments poor with hardware. In this paper, we proposed to use a model consisting of two stages: the first stage of the feature extraction process using autoencoder, and we made use of reducing the size of the features extracted from the training groups in order to reduce the size and complexity of the classification model represented by the second stage. The proposed model provided good results in comparison with the size and complexity of the classification model.