RESEARCH HIGHLIGHTS

Pattern Recognition and Machine Learning

Dimensional Reduction and Model Selection in High-dimensional Data Clustering Analysis with Applications
	Staff	Prof. CHEUNG, Yiu Ming (Principal Investigator)
	URL	http://www.comp.hkbu.edu.hk/~ymc/project/RGC_Visit_Yr11/Project_by_ymc.pdf
	Abstract	This project is to develop a single learning paradigm for three major tasks in high-dimensional data clustering analysis: (1) Dimensional Reduction, (2) Model Selection, and (3) Model Parameter Estimation. In this approach, both of the feature relevance and the cluster structure are considered. Extensive experiments show the efficacy of the proposed approach.

On Feature Selection in Gaussian Mixture Clustering
	Staff	Prof. CHEUNG, Yiu Ming (Principal Investigator)
	URL	http://www.comp.hkbu.edu.hk/~ymc/project/Feature_Selection/project.htm
	Abstract	This project proposes a new method for feature selection in Gaussian mixture clustering, and integrates it with our recently proposed RPEM algorithm so that the three closely-related issues: data clustering, model selection, and the feature selection, are all performed in a single learning process, but not a separated way. Empirical studies have shown the efficacy of the proposed approach.

On Motion Based Visual Feature Selections for Lip-Reading
	Staff	Prof. CHEUNG, Yiu Ming (Principal Investigator)
	Abstract	Lip-reading has potential attractie applications in information security, speech recognition, and so forth. This project is to develop a new visual feature representation and extraction approach to the lip localization for lip-reading. The proposed method features high accuracy of lip localization and robust performance against the shadow. We build up a prototype of lip-reading system.

Face Template Security
	Staff	Prof. YUEN, Pong Chi (Supervisor of PhD Student Mr. FENG, Yi Cheng)
	Abstract	Security and privacy concern is one of the most important issues in biometric recognition systems. Since there are intra-class variations existing in biometric templates, the traditional encryption methods which are sensitive to variations are not available to protect biometric data. Our project goal is to develop reliable face template protecting schemes, which has high security, reliable accuray and cancelability.

Human Behavior Recognition
	Staff	Prof. YUEN, Pong Chi (Supervisor of PhD Student Mr. LIU, Chang)
	Abstract	Human behavior recognition has so many applications such as visual surveillance, human-computer interfaces, content based video retrieval etc. It is a challenging research area because the dynamic human body motions have unlimited underlying representations. Our project goal is to construct discriminative underlying human action representations, make computer be aware of human action and recognize human behaviors in various scenarios effectively and efficiently.

Project in Different Research Areas: