AN ACOUSTIC APPROACH TO DRONE IDENTIFICATION USING MACHINE LEARNING
Loading...
Authors
Kenagy, Rachel M.
Subjects
MEMS sensors
acoustic vector sensors
machine learning
mel frequency cepstral coefficients
acoustic vector sensors
machine learning
mel frequency cepstral coefficients
Advisors
Martinsen, Thor
Durante Pereira Alves, Fabio
Date of Issue
2024-06
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
Drone use in contested spaces worldwide demonstrates an urgent need for fast and accurate drone detection, localization, and identification. The miniature near-resonant MEMS acoustic vector sensor developed at NPS provides drone detection and direction of arrival but cannot identify the source of a sound. The objective of this research is to augment the sensor with a real-time identification component using machine learning (ML). Data used in this work includes 350 drone and background audio files, pre-processed into one-second overlapping windows, resulting in 22,488 samples. Twenty-seven (27) features were extracted from each sample including signal and frequency statistics and mel frequency cepstral coefficients. Correlation analysis and neighborhood component analysis were used to select useful features for ML, decreasing feature dimensionality from 27 to 13. Thirty-two (32) ML algorithms were trained and tested using standard MATLAB parameters to establish a baseline performance for a variety of algorithms. Six algorithms resulted in an accuracy above 90% with acceptable prediction speed. Optimization did not result in higher accuracy or faster prediction speed. A simple neural network using the rectified linear unit (ReLu) activation function and the ensemble bagged trees algorithm are recommended for integration with the sensor based on their accuracy, speed, and potential for extension to multi-class problems.
Type
Thesis
Description
Series/Report No
Department
Organization
Identifiers
NPS Report Number
Sponsors
ONR, Arlington, VA 22203
Funding
Format
Citation
Distribution Statement
Distribution Statement A. Approved for public release: Distribution is unlimited.
Rights
This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.