A Comparative Study of CNN and CNN-GRU Models for Human Activity Recognition Using the WISDM Dataset

Human Activity Recognition (HAR) is an important field in building intelligent and context-aware systems, especially in healthcare, rehabilitation, and daily activity monitoring. This study examines the performance of two deep learning approaches on the WISDM dataset, which contains six types of activities recorded by smartphone sensors. The first approach is a Convolutional Neural Network (CNN) designed to capture spatial features from the input signals. The second approach combines CNN with a Gated Recurrent Unite (GRU) layer to better model the temporal relationships within the data. The experimental results show that the CNN-GRU model provides higher classification accuracy compared to the standalone CNN. These findings confirm that combining convolutional and recurrent structures can enhance the robustness and reliability of HAR systems, and they suggest the potential of such hybrid models for practical applications.

کلیدواژه ها:

Human Activity Recognition ، WISDM Dataset ، Convolutional Neural Network ، Gated Recurrent Unite ، Deep learning