Wavelet Packet Entropy Analysis for Detecting Humidity-induced Interference in Smart Gas Sensors

Abstract

Humidity significantly affects the accuracy and stability of smart gas sensors by altering response characteristics, introducing signal drift, and masking target gas signals, which poses challenges for reliable gas detection under variable environmental conditions. Consequently, humidity interference remains a key bottleneck for dependable gas sensing in resource-constrained embedded and wearable Internet of Things (IoT) systems. Although this study employs simulated synthetic datasets, all simulation parameters were carefully benchmarked against empirical sensor characteristics reported in existing literature, ensuring high fidelity to real-world behavior. This paper proposes a physics-informed computational framework based on Wavelet Packet Modal Entropy (WPME) to detect humidity-induced signal degradation without hardware augmentation and establishes a quadratic entropy-humidity model (WPME = 0.059A² − 0.119A + 0.094) derived from simulated NH₃ signals under 30%~86% RH. This model identifies a critical transition at 62.0% RH, which marks the shift from monolayer adsorption to capillary condensation. Below this threshold, entropy decreases by about 0.05 bits per 10% RH, while above it, the rate accelerates to 0.12 bits per 10% RH. The method achieves 98.26% detection accuracy (with only 1.47% degradation under extreme humidity) with a 0.16 ms response time, which demonstrates the viability of WPME as a fast, hardware-free solution for real-time humidity compensation in smart gas sensors.