Publication

Abstract : Infrasound sensing-based elephant monitoring has greatly aided wildlife conservation and human–elephant conflict mitigation. This research proposes an FFT-based feature approach for elephant Infrasound signal classification using supervised learning models. Classically, signals are considered primarily in raw form or time-domain features and frequency-domain features. This work moves into a new direction, applying spectral features via Fast Fourier transform (FFT) for better signal representation. A new spectral feature, Spectral Energy-RMS Ratio, is introduced to improve classification. The supervised models are trained to classify elephant and non-elephant Infrasound signals after being subjected to pre-processing and feature aggregation, the models achieved 91.79% accuracy. Experimental results prove a better accuracy rate than the baseline systems, highlighting the novelty and efficiency of FFT-based feature extraction for the reliable classification of elephant Infrasound signals.