Research Grants


The Big Australian Speech Corpus: An audio‑visual speech corpus of Australian English

CIs: Prof Denis K Burnham, Dr Felicity M Cox, Prof Andrew R Butcher, A/Prof Janet M Fletcher, Prof Michael Wagner, Dr Julien R Epps, Dr John C Ingram, Dr Joanne Arciuli, Dr Roberto Togneri, Dr Philip J Rose, Dr Nenagh M Kemp, Prof Anne Cutler, Prof Robert Dale, Dr Takaaki Kuratate, Prof David M Powers, A/Prof Stephen Cassidy, Dr David B Grayden, Dr Deborah E Loakes, Prof Dr Mohammed Bennamoun, Dr Trent W Lewis, Dr Roland Goecke, Prof Catherine T Best, A/Prof Steven Bird, Prof Eliathamby Ambikairajah, Prof John T Hajek, Dr Shunichi Ishihara, Dr Yuko Kinoshita, Dr Dat T Tran, Dr Girija Chetty, Prof Mark Onslow

Grant: ARC LIEF (ID: LE100100211, AUD $650,000.00)

Abstract: Contemporary speech science and technology are driven by the availability of large speech corpora. While audio databases exist for languages spoken in America, Europe and Japan, there is currently no large auditory‑visual database of spoken language, and certainly not one for Australian English. Here we will establish the Big Australian Speech Corpus, which will support a speech science research and development using Australian English and facilitate the development of Australian speech technology applications from automatic speech recognition and text‑to‑speech synthesis used in taxi and other ordering services, to hearing prostheses and talking head aids for learning‑impaired children, and a range of security and forensic applications




Knowledge discovery in hospital pathology databases


CIs: Prof Simon Hawkins, Dr Alice Richardson, Dr Brett Lidbury

AIs: A/Prof Dharmendra Sharma, Mr Gus Koerbin, Prof John Fulcher, Dr Dat Tran

Grant: University Interdisciplinary Research Grant

Abstract: In today’s health care environment, the amount of electronically available clinical pathology data has grown as hospital information systems become more commonplace. This data can provide the basis for analyzing risk factors for many diseases. This project will apply data mining techniques to clinical pathology databases to discover new numerical rules that will identify small groups in the population that are at high relative risk of certain viral diseases. Once rules have been discovered, they can be used to develop a system of electronic alerts that will identify patterns of pathology results that place a patient at a high relative risk of a viral disease.  A secondary aim of the project is to enrich the established and thriving UC research program in human virology



Fuzzy Pattern Recognition Methods for Intrusion Detection and Spam Emails Filtering Systems

CIs: Dr Dat Tran, Dr Wanli Ma, A/Prof Dharmendra Sharma, Dr Shuangzhe Liu

Grant: BLIS Research Grant

Abstract: We propose fuzzy pattern recognition methods for intrusion detection and spam emails filtering systems.  The project aims to build up reliable intrusion detection spam emails filtering systems for activities over computer networks and telephone lines. These systems will be parts of a multi-agent system which is developing at School of Information Sciences and Engineering. The proposed methods including fuzzy hidden Markov modelling, fuzzy normalisation, fuzzy Gaussian mixture modelling and fuzzy entropy clustering that have shown good performance in biometric authentication and bioinformatics. The systems have wide applications in e-commerce, defence, human society and computer security operations, high-tech crime investigation, and current battle against terrorism.

Intrusion Detection with Temporal Compression on Network Traffic Records

CIs: Dr Wanli Ma, Dr Dat Tran

Grant: BLIS Research Grant

Abstract: An intrusion detection system monitors the computer network for possible intrusions. There are many proposals, yet the simple requirement of being efficient and effective is far from achievable. Most proposed IDS ignored the time context when interpreting the network traffic record data. We propose to interpret the data in the context of time. We propose 2 approaches: (1) using data aggregation to associate time information and clustering engine of data mining for intrusion detection; (2) threading network traffic records based on network session activities along the time axis and using (Hidden) Markov Model for intrusion detection. The former will be more efficient to perform but less accurate, and vice versa for the later. Both should yield better detection rate. An intrusion detection system is an important for ICT security. We have been working in this area for some times and laid a broad foundation. Our next step is to build the depth and produce quality research outcomes. This research work is also in line with Australia National Research Priorities – Safeguarding Australia.



Statistical and Fuzzy Techniques for Classification of Cell Nuclei in Different Mitotic Phases

CI: Dr Dat Tran

Grant: BLIS Completion Research Grant

Abstract: This project aims to develop an innovative and comprehensive application of statistical and fuzzy pattern recognition for the computerized classification of cell nuclei in different mitotic phases. We are interested in applying several advanced computational, probabilistic, and fuzzy-set methods we have proposed in speech, speaker and image recognition for the computerized classification of cell nuclei in different mitotic phases.

New Statistical Modelling Methods for Biometric Authentication and Intrusion Detection Systems

CIs: Dr Dat Tran, A/Prof Dharmendra Sharma, Dr Wanli Ma, Dr Shuangzhe Liu

Grant: University Multidisciplinary Research Grant

Abstract: We propose new statistical modelling methods that can be applied to biometric authentication, intrusion detection and bioinformatics systems.  The project aims to build up a reliable authentication and intrusion detection system for activities over computer network, telephone lines, and closed-circuit TV (CCTV) footages. The proposed methods including quasi-likelihood estimation, temporal hidden Markov models, and background modelling-based authentication can avoid limitations of current systems and therefore enhance the system performance and accuracy. The system has a wide application in e-commerce, defence, human society and computer security operations, high-tech crime investigation, and current battle against terrorism.

A Multiagent Based IT Security Framework

CIs: A/Prof. Dharmendra Sharma, Dr. Wanli Ma, Dr. Dat Tran, and Dr. Shuangzhe Liu,

Grant: BLIS Completion Research Grant


Maximum likelihood and quasi likelihood estimation methods and application

CIs: Dr. Shuangzhe Liu, A/Prof. Dharmendra Sharma, Dr. Dat Tran, Dr. Wanli Ma

Grant: BLIS Completion Research Grant


Some Likelihood Related Issues, Optimal Estimation and Inference with Applications

CIs: Dr Shuangzhe Liu, Dr Dat Tran, Dr Wanli Ma, A/Prof Dharmendra Sharma

Grant: BLIS Research Grant

Abstract: In this project, we shall study likelihood related ideas, methods and other data mining techniques. We shall obtain efficiency comparison and model selection results with applications to pattern recognition, classification and evidence specification and interpretation. The concept of likelihood plays an essential role in data mining and statistics. Yet it is only recently recognized by researchers for several areas. For example ideas and methods based on likelihood ratio and quasi likelihood and their applications to speaker recognition and evidence specification are still at an early but promising stage of research. The project is significant also in the sense that the issues addressed are so important and multidisciplinary of statistics, IT, forensics and security issues, and are shared in the research areas of colleagues across disciplines and schools at the University



Person Authentication Using Speaker Verification and Verbal Information Verification

CI: Dr Dat Tran

Grant: Divisional Research Institute Grant

Abstract: We propose a new method for person authentication that can achieve a considerably higher level of security. The voice characteristics and the information content of spoken phrases will be investigated for person authentication by combining speaker verification (SV) and verbal information verification (VIV). VIV is the process of verifying spoken utterances against personal information, such as the date of birth or the mother’s maiden name, which is stored in a given personal data profile. The system involves two phases, enrolment and verification, which are described in the following subsections

    Enrolment phase: A key code, such as an account number, is assigned to each ”client” of the system. The client is then asked to provide a set of personal information items, such as date of birth, address, home telephone number, etc. A microphone is set up to collect the speech waveforms. Speech data after feature extraction are used to train speaker hidden Markov models (HMMs) for the SV sub-system and sub-word HMMs for the VIV sub-system.

     Verification phase: An identity claim is made by an unknown person. The speech data obtained from the unknown person are sent to the SV sub-system to verify the speaker and to the VIV sub-system to verify the content of the spoken utterance. The results from the two sub-systems are similarity scores, which are fused and compared with a given threshold to either accept or reject the unknown person’s identity claim



A combined ensemble-fuzzy normalisation method for speaker verification

CI: Dr Dat Tran

Grant: ARC Small Grant

Abstract: The project investigates novel ensemble methods in machine learning and fuzzy pattern recognition approaches to speech and speaker recognition. The aims are the further developments of Bagging and AdaBoost methods and fuzzy hidden Markov models to speech and speaker recognition systems. AdaBoost by re-weighting using fuzzy hidden Markov models is also the alternative aim of the project. The expected outcomes of the research project are new leading-edge methodologies for speech and speaker recognition using ensemble methods and fuzzy hidden Markov models. The significance of the project is the combined fuzzy and ensemble methodology has the potential to develop into a new technology for speech and speaker recognition.