Machine Learning-Based Classification of Mango Pulp Weevil Activity Utilizing an Acoustic Sensor

Area of Science:

• Agricultural Entomology
• Acoustic Signal Processing
• Machine Learning Applications

Background:

• The mango pulp weevil (MPW) is a significant agricultural pest causing substantial economic losses due to difficulties in detection.
• Current detection methods for MPW are inadequate as the pest leaves no external physical damage signs on mangoes.
• Palawan is under quarantine for mango exports following the discovery of MPW, highlighting the urgent need for effective detection strategies.

Purpose of the Study:

• To develop a non-invasive method for detecting the mango pulp weevil (MPW) using audio feature extraction and machine learning.
• To evaluate the efficacy of acoustic sensors for identifying MPW activities.
• To establish a baseline for future advancements in automated MPW detection systems.

Main Methods:

• Utilized MATLAB machine learning tools for audio feature extraction and classification.
• Evaluated the performance of different acoustic sensors, selecting the MEMS sensor for its optimal results and accessibility.
• Recorded and analyzed acoustic data from MPW activities (walking, resting, mating) within a soundproof chamber.
• Employed Mel-frequency cepstral coefficient (MFCC) for feature extraction and Support Vector Machine (SVM) for classifier training.

Main Results:

• The MEMS acoustic sensor demonstrated superior performance in capturing MPW-generated sounds.
• The study successfully identified distinct acoustic signatures for adult MPW activities including walking, resting, and mating.
• Achieved an overall accuracy of 89.81% in characterizing mango pulp weevil activities through audio analysis and machine learning.

Conclusions:

• Non-invasive audio feature extraction combined with machine learning offers a viable method for detecting the mango pulp weevil (MPW).
• The developed acoustic detection system shows significant potential for improving pest management strategies in mango cultivation.
• This research provides a foundation for developing advanced, automated systems for early MPW detection, reducing crop losses and trade restrictions.