"RPW Acoustic Data"

"Early stage monitoring of Red Palm Weevil (RPW) activity in palms  is essential to prevent severe structural damage and economic losses in palm cultivation. Unlike conventional inspection methods, acoustic-based detection offers a non-invasive solution but is challenged by background noise and subtle signal variations. The proposed work uses an adaptive statistical signal processing framework for RPW detection.The method integrated i) adaptive sliding window segmentation with overlapping analysis to capture temporal variations, ii) RMS and auto-correlation descriptors combined with spectra-temporal features from a Short-Time Fourier Transform (STFT) and iii) Statistical validation using a two sample t-test to confirm significant differences (p<0.01) between RPW and Non RPW acoustic traits.The extracted features are further enhanced using a Gaussian-weighted spectrogram representation before classification. Among multiple classifiers, the lightweight MobileNetV2 model achieved the highest accuracy (97.88%), with an  AUC of 0.96 and an F1 score of 0.888, outperforming SVM, KNN, CNN, and RNN.The complete pipeline is deployed on a Raspberry Pi- based low power acoustic sensing unit, demonstrating real time on device feature extraction and classification, making the system suitable for continuous in field pest surveillance. This approach provides an eco-friendly and computationally efficient framework for real-time pest monitoring and contributes to advancing signal processing techniques for biological acoustic detection."