Through-Wall Detection using Software-Defined Radio based on adaptive Principal Component Analysis

Through-Wall Detection (TWD) using opportunistic WiFi signals enables non-invasive sensing for security and rescue applications; however many existing approaches rely on controlled access points or specialised hardware. This paper presents a TWD system that extracts Channel State Information (CSI) from ambient WiFi packets using a customised software-defined radio (SDR), Bluebottle, without requiring transmitter control. The key contribution is a spectral-domain scoring mechanism for adaptively selecting motion-relevant principal components from a Principal Component Analysis (PCA) decomposition of the CSI, using Welch power spectral density estimates to quantify each component's signal-to-noise ratio and spectral concentration within the frequency band associated with human motion. The selected components are then analysed using a continuous wavelet transform to robustly identify time-localised motion events. Experimental results demonstrate that the proposed adaptive component-selection method consistently reduces false detections and produces sharper time-frequency energy ridges compared to conventional fixed-component PCA.