Public Dataset for Simultaneous Human Activity Recognition and Localization using WiFi Signals

WiFi-based human sensing technology has unique properties such as widespread availability, resistance to light interference, overcoming terrain limitations, and respecting user privacy. It has been utilized in various fields. One interesting application is simultaneous human activity recognition and localization. However, CSI dataset collection is labor-intensive and affected by hardware requirements and timing synchronization between transmitters. Therefore, datasets are not publicly released in some studies. Other studies used a cable for timing between transmitters which limits the practicality in real-world scenarios. In this study, we release the first public dataset with wireless timing synchronization, WiARL-UIT, for simultaneous human activity recognition and localization. Finally, we have evaluated the performance of various machine learning models. As a result, the best accuracy is 97.0% when a convolutional neural network (CNN) model is used, showing the feasibility of the research direction of simultaneous human activity recognition and localization. The dataset and source code are publicly shared at https://github.com/F4tt/WiARL-UIT.