Audio Zooming in a Drone Surveillance System for Police Evidence Gathering

Stroud, S ORCID: 0009-0005-3684-5050 (2026) Audio Zooming in a Drone Surveillance System for Police Evidence Gathering. Doctoral thesis, Liverpool John Moores University.

Preview

Text 2026stroudphd.pdf.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (13MB) | Preview

Abstract

This thesis documents the work on an innovative audio-zooming system intended to be attached to and then deployed on a drone, which is then able to be used for Police surveillance and evidence gathering. The central hypothesis is that a lightweight array of omni-directional microphones, when integrated with an advanced beamforming algorithm, noise reduction, and filtering techniques, can effectively detect and target noise, then separate and isolate the desired sound sources. This capability could significantly enhance the collection of auditory evidence from complex and noisy environments. This, in turn, could support the Police in law enforcement efforts to secure reliable convictions via forensic evidence-gathering methods.
The thesis begins with an introduction to the historical challenges associated with the Cocktail Party Problem, a term describing the human ability to detect and focus on a single sound source in a noisy environment with competing, interfering sources, and the longstanding attempts of the scientific community to replicate this biological phenomenon through engineering solutions. A literature review follows, exploring the multi-disciplinary aspects of sound localisation and sound separation problems, covering advancements in signal processing, machine learning, and acoustic engineering. Prior to the experimental approaches being covered, the acoustic-physics and mathematical foundations underpinning beamforming and noise-suppression theory are explored. Building on that groundwork, there is a focus on the algorithms used for directional sound capture, such as the Minimum Variance Distortionless Response (MVDR) beamformer, together with advanced audio techniques such as spectral filtering to improve legibility. The discussion proceeds to the equipment that make the experiments possible. MATLAB® serves as a flexible environment for simulation, while a set of custom-built microphone arrays gathers the raw acoustic data. An examination of the experimental approach rounds things out by explaining, step by step, how the listening trials were designed and executed, including array configurations, test-room conditions, and the data-analysis pipeline that converts recordings into interpretable results. The results from the experiments are presented and analysed to evaluate the performance of the proposed system, including metrics on signal-to-noise ratio improvements, directional accuracy, and computational efficiency.
The conclusion summarises the findings, affirming the potential and limitations of the proposed system while suggesting pathways for future research, such as integrating real-time processing capabilities and testing in varied environmental conditions. This thesis contributes to the growing field of audio and video engineering by offering a novel framework for drone-mounted auditory surveillance, opening avenues for enhanced law enforcement and forensic practices.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Audio Zooming, Surveillance, Forensic Evidence Gathering, Beamforming, Digital Signal Processing
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T58.5 Information Technology H Social Sciences > HV Social pathology. Social and public welfare. Criminology > HV7231 Criminal Justice Administrations > HV7551 Police. Detectives. Constabulary
Divisions:	Engineering
Date of acceptance:	26 February 2026
Date of first compliant Open Access:	6 March 2026
Date Deposited:	06 Mar 2026 10:55
Last Modified:	06 Mar 2026 10:55
DOI or ID number:	10.24377/LJMU.t.00028160
Supervisors:	Jones, K, Edwards, G, Ellis, D and Robinson, C
URI:	https://researchonline.ljmu.ac.uk/id/eprint/28160

View Item