Towards a Low-cost WiFi based Real-time Human Activity Recognition System

Current implementations of human monitoring systems based on video, audio, and wearables offer better data but at the cost of privacy and convenience. While research has focused on systems using off-the-shelf WiFi hardware as an alternative to existing systems, most of them have been implemented using the Intel WL5300 WiFi network adapter, which requires a dedicated computer to function. Our research focuses on using low-cost Raspberry Pi 3B+ devices as an alternative for human activity recognition using WiFi CSI data through classification. In this paper, we propose a real-time implementation of a deep learning based human activity recognition system through classification using Raspberry Pi. We have created a public dataset of human activity data for six activities. A Convolutional LSTM model is used for the classification of activity data. A prototype system has also been developed for the real-time recognition of human activity data. We have achieved an accuracy of 95% for the model for the experiments performed in two test environments across six activities, including one for no movement. We have also evaluated the performance of our real-time human activity recognition system with acceptable performance in a static environment.