Real-Time CSI-Based Wireless Gesture Recognition for Human-Machine Interaction

The study and the design of novel methodologies and techniques for user's activity and gesture recognition is of great interest and a hot topic in human-computer interactions. Hand gesture recognition techniques based on computer-vision have yielded impressive results, but they involve users’ privacy concerns, therefore other sensing approaches are of interest. In this work, a novel machine learning methodology based on passive electromagnetic sensing that exploits commodity Wi-Fi signals is proposed. Such an approach has been preliminary validated in a real house environment with a classification accuracy of 98%.