Smartphone-based CNN-GRU Framework for Human Activity Recognition

Human activity recognition has been a widely explored topic for the past several years due to its application in mankind such as smart homes, health assistance, intrusion detection, behavioral recognition, and so on. Deep learning has been widely exploited for recognition of human activities especially CNN and its hybridization with Recurrent Neural Network (RNN). In this article, a hybrid deep learning model with the combination of Convolutional Neural Network and Gated Recurrent Unit has been proposed for human activity recognition. UCI-HAR smartphone dataset is taken for performance evaluation of this proposed model and it is found that the framework performed very well with the mean accuracy of 92.53%.