WiFi CSI-Based Earthquake Early Warning System Using Edge Analytics
Author : Dr. K Karuppasamy, S Anuba, M Vijayasharathi, A Sachin, S Vijesh and S Sowmya
Abstract :
The rapid advancements in wireless technologies have paved the way for utilizing Radio Frequency (RF) signals for intricate environmental sensing. While traditionally relied upon for human activity recognition, this paper proposes repurposing ubiquitous WiFi Channel State Information (CSI) for a low-cost, highly scalable Earthquake Early Warning (EEW) system. Conventional EEW systems depend on specialized, expensive seismic sensors, limiting their deployment in developing regions. In contrast, our proposed architecture leverages the fine-grained micro-vibration sensitivity of standard MIMO-OFDM WiFi signals in the 0.3–5 Hz frequency range to detect primary P-wave signatures. Extracted via the Nexmon firmware on a Raspberry Pi, the raw CSI data undergoes rigorous signal processing, including wavelet-based denoising and feature extraction. To handle the high-dimensional spatial-temporal data, a hybrid Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) deep learning model is deployed directly at the edge, ensuring ultra-low latency inference without cloud dependency. Furthermore, a LoRa-based backup communication module is integrated to guarantee alert transmission during catastrophic cellular network failures. Experimental evaluations demonstrate a detection accuracy of 94.3% with an average alert latency of 1.4 seconds. The proposed architecture shifts the paradigm of disaster management toward ubiquitous, low-cost IoT sensing.
Keywords :
Channel State Information (CSI), Deep Learning, Earthquake Early Warning, Edge AI, Internet of Things, LoRa, Signal Processing, WiFi Sensing.
