Fusion-Driven Acoustic Intelligence for Insect Detection in Grain Storage

Abstract

Insect infestations in stored grains continue to pose a serious threat to food quality, safety, and supply, often leading to significant post-harvest losses. Studies show that insects alone contribute to more than 10% of the total post-harvest losses in grains and cereals. Traditional detection methods are not only time-consuming and labor-intensive but also prone to inaccuracies. To address these challenges, this study presents a noninvasive and scalable approach that combines acoustic signal analysis with feature fusion techniques. By analyzing insect-generated sounds from three standard datasets, the system captures movement and feeding activity. A fusion of spectral, cepstral, and statistical features enhances detection performance, while a phase-based speech enhancement method helps reduce background noise for clearer signal interpretation. These features are then used with standard audio classification models to determine insect presence and activity levels. Experimental results show the method achieves an average detection accuracy of 94%. Designed to be both practical and efficient, this solution offers a reliable way to protect stored grains and reduce losses across large-scale storage facilities.