A Lightweight Signal-Processing Front-End for Acoustic Forest Fire Detection
Year: 2026
Type: conference-paper
Abstract: This paper formulates a lightweight signal-processing front-end for acoustic forest fire detection as a reusable methodological layer within a broader audio-analysis pipeline. The study consolidates evidence from eight thematic literature sets covering feature fusion, autocorrelation, handcrafted and interpretable descriptors, Hilbert-envelope analysis, low-complexity audio descriptors, spectral representations, and time-frequency modeling, and relates this corpus to three prior fire/non-fire studies developed by the present research group. The main contributions are a structured synthesis of the relevant descriptor literature, a lightweight and classifier-agnostic front-end design, and a compact mathematical formulation suitable for downstream integration with PCA-based models, conventional machine-learning classifiers, CNNs, RNNs, Audio Spectrogram Transformers, and hybrid CNN–Transformer architectures. The resulting manuscript provides the theoretical and methodological foundation required to position the proposed front-end as a reusable component for acoustic fire monitoring.
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7170925569