Acoustic Sensing and Machine Learning for Early Detection of Insect Activity in Stored Grains
Journal of Smart and Sustainable Farming · 22.10.2026 · 239
DOI: 10.64026/jssf/2026020 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Smart and Sustainable Farming, online erschienen am 22.10.2026. Verfasst von Nina Cedergreen Cedergreen und 1 weiteren Autor:innen. Die Arbeit nennt 14 Literaturangaben.
Abstract (Originalsprache)
Insects hidden inside stored grain are difficult to detect before damage becomes economically significant, and conventional sampling and trapping are slow, labour-intensive and insensitive at low insect densities. This article presents an acoustic sensing and machine-learning framework for early detection of insect activity in grain storage. The framework combines band-limited filtering and spectral subtraction with a fused vector of 113 descriptors (MFCC statistics, log-mel statistics, spectral-shape measures and pulse-train timing features) and a lightweight classifier that separates clean grain from three insect classes. No labelled field corpus was available for this study, so the framework is evaluated on a physically motivated synthetic benchmark in which insect feeding and movement are modelled as trains of damped sinusoidal pulses embedded in grain-bin noise with impulsive distractors. All numerical results are therefore simulated and should be read as a controlled comparison of methods, not as field performance. Under five-fold cross-validation at clip-level signal-to-noise ratios (SNR) of −24 to −4 dB, the proposed pipeline reached 87.9% four-class accuracy against 78.8% for an MFCC-only baseline, and detected insect activity with an F1 score of 0.961. In an independent noise sweep, the proposed pipeline outperformed the baseline by 8.5 to 14.5 percentage points between −24 and −12 dB. Feature extraction took about 15 ms per two-second clip on one CPU core. Limitations of simulation and the field validation required are discussed.
Zitieren
APA
Cedergreen, N. C., & Lemmi, S. O. (2026). Acoustic Sensing and Machine Learning for Early Detection of Insect Activity in Stored Grains. Journal of Smart and Sustainable Farming, 239. https://doi.org/10.64026/jssf/2026020
Vancouver
Cedergreen NC, Lemmi SO. Acoustic Sensing and Machine Learning for Early Detection of Insect Activity in Stored Grains. Journal of Smart and Sustainable Farming. 2026:239. doi:10.64026/jssf/2026020
BibTeX
@article{Cedergreen2026,
author = {Cedergreen, Nina Cedergreen and Lemmi, Sangi Olani},
title = {Acoustic Sensing and Machine Learning for Early Detection of Insect Activity in Stored Grains},
journal = {Journal of Smart and Sustainable Farming},
year = {2026},
pages = {239},
doi = {10.64026/jssf/2026020}
}
EUVIMED