Acoustic Detection and Classification of Human Postures

Qu, Yuanying (2024) Acoustic Detection and Classification of Human Postures. PhD thesis, University of Liverpool.

Text 201254213_Aug2024.pdf - Author Accepted Manuscript Access to this file is embargoed until 1 January 2030. Download (13MB)

Abstract

As advanced human detection and recognition technologies, acoustic detection systems offer broad application prospects in security monitoring, emergency response, and environmental sensing. With the continuous advancement of biometric recognition technology, acoustic methods have gained attention due to their alignment with natural biometric characteristics. However, current systems face real-time performance and accuracy challenges, especially in multi-person detection. This thesis explores the potentials of acoustic detection and classification of human walking postures by investigating the principles, methods, and applications. The study analyses footstep sound features to classify and recognise walking postures, which reveals a significant correlation between Doppler shifts and footstep vibration characteristics. This correlation enhances the effectiveness and accuracy of human recognition and detection. A successful multi-person detection and recognition method in complex scenarios with simultaneous walkers is proposed in this thesis. A novel deep learning-based acoustic detection system for multi-person detection and recognition has also been developed. Based on the multi-person research, further the correlation feature is utilised, and the deep learning model is optimised to improve the performance of the acoustic multi-person detection system. Key contributions include exploring the relationship between acoustic signals and human biometrics, achieving multi-person detection in complex environments, and developing an integrated system. This system offers significant potentials for human detection and classification applications, providing new insights and methodologies for research and practical implementations. Future work will optimise system performance and explore broader application prospects.

Item Type:	Thesis (PhD)
Divisions:	Faculty of Science and Engineering Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	21 Jan 2025 10:43
Last Modified:	03 Feb 2025 11:40
DOI:	10.17638/03187285
Supervisors:	Wang, Xinheng ORCID: 0000-0001-8771-8901 Zhou, Jiafeng
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3187285