Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms

Daniel White; Mohammed Jahangir; Chris Baker; Michail Antoniou

doi:10.1109/TRS.2023.3326317

Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms

Daniel White, Mohammed Jahangir^*, Chris Baker, Michail Antoniou^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

78 Downloads (Pure)

Abstract

In this article, a method for creating highly realistic synthetic drone micro-Doppler spectrograms is presented and its effectiveness of training a bird-drone classifier for real scenario classification is shown via comparisons to a real benchmark. The effect of drone motor speed sampling used when simulating drone micro-Doppler is shown to have a significant impact on the accuracy of synthetic results and variations of this approach are explored. Four synthetic datasets were created differing in motor speed sampling and each were compared in their ability to train a convolutional neural network to classify real data. The highest fidelity synthetic dataset achieved a classification accuracy of 86.6% compared to the real benchmark accuracy of 89.7%. The adverse effect on classifier robustness when reducing the simulation fidelity by altering the motor speed sampling is shown.

Original language	English
Number of pages	14
Journal	IEEE Transactions on Radar Systems
Early online date	20 Oct 2023
DOIs	https://doi.org/10.1109/TRS.2023.3326317
Publication status	E-pub ahead of print - 20 Oct 2023

Access to Document

10.1109/TRS.2023.3326317Licence: None: All rights reserved

WhiteD2023Urban_AAM
This is the Accepted Author Manuscript of the following article: D. White, M. Jahangir, C. J. Baker and M. Antoniou, "Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms," in IEEE Transactions on Radar Systems, doi: 10.1109/TRS.2023.3326317. © 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
Accepted author manuscript, 4.71 MBLicence: Other (please specify with Rights Statement)

MEFA: Mapping and Enabling Future Airspace
Reynolds, J. (Co-Investigator), Antoniou, M. (Principal Investigator), Sadler, J. (Co-Investigator), Baker, C. (Co-Investigator) & Baker, C. (Co-Investigator)
Engineering & Physical Science Research Council
1/04/20 → 31/03/24
Project: Research Councils
UK National Quantum Technology Hub in Sensing and Timing
Attallah, M. (Co-Investigator), Jones, R. (Co-Investigator), Metje, N. (Co-Investigator), Constantinou, C. (Co-Investigator), Faramarzi, A. (Co-Investigator), Singh, Y. (Co-Investigator), Holynski, M. (Principal Investigator), Bongs, K. (Co-Investigator), Baker, C. (Co-Investigator), Antoniou, M. (Co-Investigator) & Stewart, E. (Co-Investigator)
Engineering & Physical Science Research Council
1/12/19 → 30/11/24
Project: Research Councils

Cite this

@article{f480eb20a06c495999870e7435939fff,

title = "Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms",

abstract = "In this article, a method for creating highly realistic synthetic drone micro-Doppler spectrograms is presented and its effectiveness of training a bird-drone classifier for real scenario classification is shown via comparisons to a real benchmark. The effect of drone motor speed sampling used when simulating drone micro-Doppler is shown to have a significant impact on the accuracy of synthetic results and variations of this approach are explored. Four synthetic datasets were created differing in motor speed sampling and each were compared in their ability to train a convolutional neural network to classify real data. The highest fidelity synthetic dataset achieved a classification accuracy of 86.6% compared to the real benchmark accuracy of 89.7%. The adverse effect on classifier robustness when reducing the simulation fidelity by altering the motor speed sampling is shown.",

author = "Daniel White and Mohammed Jahangir and Chris Baker and Michail Antoniou",

year = "2023",

month = oct,

day = "20",

doi = "10.1109/TRS.2023.3326317",

language = "English",

journal = "IEEE Transactions on Radar Systems",

}

TY - JOUR

T1 - Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms

AU - White, Daniel

AU - Jahangir, Mohammed

AU - Baker, Chris

AU - Antoniou, Michail

PY - 2023/10/20

Y1 - 2023/10/20

N2 - In this article, a method for creating highly realistic synthetic drone micro-Doppler spectrograms is presented and its effectiveness of training a bird-drone classifier for real scenario classification is shown via comparisons to a real benchmark. The effect of drone motor speed sampling used when simulating drone micro-Doppler is shown to have a significant impact on the accuracy of synthetic results and variations of this approach are explored. Four synthetic datasets were created differing in motor speed sampling and each were compared in their ability to train a convolutional neural network to classify real data. The highest fidelity synthetic dataset achieved a classification accuracy of 86.6% compared to the real benchmark accuracy of 89.7%. The adverse effect on classifier robustness when reducing the simulation fidelity by altering the motor speed sampling is shown.

AB - In this article, a method for creating highly realistic synthetic drone micro-Doppler spectrograms is presented and its effectiveness of training a bird-drone classifier for real scenario classification is shown via comparisons to a real benchmark. The effect of drone motor speed sampling used when simulating drone micro-Doppler is shown to have a significant impact on the accuracy of synthetic results and variations of this approach are explored. Four synthetic datasets were created differing in motor speed sampling and each were compared in their ability to train a convolutional neural network to classify real data. The highest fidelity synthetic dataset achieved a classification accuracy of 86.6% compared to the real benchmark accuracy of 89.7%. The adverse effect on classifier robustness when reducing the simulation fidelity by altering the motor speed sampling is shown.

U2 - 10.1109/TRS.2023.3326317

DO - 10.1109/TRS.2023.3326317

M3 - Article

JO - IEEE Transactions on Radar Systems

JF - IEEE Transactions on Radar Systems

ER -

Urban Bird-Drone Classification with Synthetic Micro-Doppler Spectrograms

Abstract

Access to Document

Fingerprint

Projects

MEFA: Mapping and Enabling Future Airspace

UK National Quantum Technology Hub in Sensing and Timing

Cite this