Online Audio-Visual Multi-Source Tracking and Separation: A
Labeled Random Finite Set Approach

Ong, Jonah Soon Xuan

Ong JSX 2021 Public.pdf (10.66Mb)

Access Status

Open access

Authors

Ong, Jonah Soon Xuan

Date

2021

Supervisor

Ba Tuong Vo

Sven Nordholm

Ba-Ngu Vo

Type

Thesis

Award

PhD

Metadata

Show full item record

Faculty

Science and Engineering

School

School of Electrical Engineering, Computing and Mathematical Sciences

URI

http://hdl.handle.net/20.500.11937/89300

Collection

Curtin Theses

Abstract

The dissertation proposes an online solution for separating an unknown and time-varying number of moving sources using audio and visual data. The random finite set framework is used for the modeling and fusion of audio and visual data. This enables an online tracking algorithm to estimate the source positions and identities for each time point. With this information, a set of beamformers can be designed to separate each desired source and suppress the interfering sources.