Robust Acoustic Feature Selection for Embedded Detection of Queenless Honey Bee Colonies
Supervisor Name
Emad Natsheh
Supervisor Email
e.natsheh@najah.edu
University
An-Najah National University
Research field
Artificial Intelligence
Bio
Dr. Emad Natsheh is an Associate Professor in the Faculty of Engineering and Information Technology at An-Najah National University. He holds a PhD in Artificial Intelligence and Renewable Energy and an MSc in Computer and Network Technology from Manchester Metropolitan University (UK), and a BSc in Computer Engineering from An-Najah National University. His research focuses on artificial intelligence, machine learning, smart systems, and renewable energy applications. Dr. Natsheh has published several papers in international journals and conferences and has experience in training in data science, Python programming, and data analysis.
Description
This project aims to develop and evaluate an acoustic monitoring approach for distinguishing queenright from queenless honey bee colonies. Differences between colonies, recording conditions, and environmental noise can affect classification reliability, while class imbalance may reduce the detection of queenless colonies. Building on an existing ESP32-based prototype, the research will investigate a feature-selection method that jointly considers classification performance, stability across colonies and noise conditions, and computational requirements. A selected subset of the 2022 UrBAN dataset will be used to extract mel-filterbank energies, Mel-Frequency Cepstral Coefficients (MFCCs), and spectral descriptors. CNN, SVM, and Random Forest classifiers will be compared, with training, validation, and testing separated by colony to prevent data leakage. The proposed contribution is to determine whether the selected compact features improve queenless-colony detection and robustness compared with conventional feature sets and the existing MFE-based CNN. Evaluation will include controlled noise experiments, feature ablation studies, and measurements of balanced accuracy, macro F1-score, queenless precision and recall, and false-positive rate. A suitable model will be quantized and evaluated on an ESP32 connected to an INMP441 microphone. Labeled recordings from Palestinian apiaries will support additional evaluation if available.