WiFi CSI-Based Long-Range Through-Wall Human Activity Recognition with the ESP32

Abstract

WiFi Channel State Information (CSI)-based human activity recognition (HAR) is an unobtrusive method for contactless, long-range sensing in spatially constrained environments while preserving visual privacy. Despite the presence of numerous WiFi-enabled devices around us, few expose CSI to users, resulting in a lack of sensing hardware options. Recently, variants of the Espressif ESP32 have emerged as potential low-cost, easy-to-deploy solutions for WiFi CSI-based HAR. In this work, we evaluate the ESP32-S3’s long-range through-wall HAR capabilities by combining it with a 2.4GHz directional biquad antenna. The experimental setup uses a transmitter-receiver configuration spanning 18.5m across five rooms. We assess line-of-sight (LOS) and non-line-of-sight (NLOS) performance using CNN HAR models trained on CSI spectrograms. CSI HAR datasets used in this work consist of 392 LOS and 384 NLOS spectrograms from three activity classes and are made publicly available. The gathered results clearly demonstrate the feasibility of long-range through-wall presence detection and activity recognition with the proposed setup.