WiFi Sensing for Human Activity Recognition

Abstract:

This talk presents recent advances in WiFi sensing for human activity recognition (HAR), demonstrating how existing wireless communication infrastructure can be leveraged as a powerful sensing modality. By exploiting Channel State Information (CSI) readily available in commodity WiFi devices, it is possible to infer both large-scale and fine-grained human activities without requiring wearable sensors or dedicated hardware. The presentation begins with an overview of the emerging paradigm of Integrated Sensing and Communication (ISAC), including developments such as IEEE 802.11bf and future 6G systems, where communication signals are repurposed for environmental sensing. It then introduces signal processing techniques for extracting meaningful features from CSI, including time-frequency analysis and Doppler-based representations that capture motion dynamics. A key contribution discussed in the talk is the use of lightweight and scalable machine learning approaches, such as random convolutional kernels and deep neural networks, for efficient end-to-end activity <a href="http://recognition.

Speaker(s):” target=”_blank” title=”recognition.

Speaker(s):”>recognition.

Speaker(s): Professor Shahrokh Valaee,

Room: EV003-309, Bldg: Electrical & Computer Engineering Department EV, Concordia University, 1515 Ste. Catherine West, MONTREAL, Quebec, Canada, H3G 1M8