Projects

Active Learning

We propose to use active learning to improve the efficiency of hidden annotation and thus improve the performance of 3D model retrieval. 

Model Updating

We propose to update the eigenspace model in order to adaptively learn the changing appearance of a object.

Audio-Visual Speech Processing

We use the video camera to track the lip movements of the user and then utilize such information to enhance the performance of acoustic speech recognition.

NetICE

We provide a system that will reproduce an actual conference setting, so that people can communicate from anywhere at any time and still feel that they are in the same environment.

Decoder Error Concealment

Error concealment is a way to recover or conceal the loss information due to the transmission errors. We develop a  UMPC-based error concealment method achieves this goal.

Optimal Watermark Detection

Watermark or data hiding techniques has a broad range of usages. We build a framework for watermark detection and develop an optimal watermark detector.

Face Authentication

We propose a new algorithm to build a face authentication system tolerant to facial expression variations and registration errors.

Progressive 3D Model

We propose a joint geometry/texture codec for the 3D models rendered with textures, resulting in progressive bitstreams.

Face Tracking

We want to build a robust and accurate face tracking system running in real-time. Also we want to extend current single face tracking system to track multiple faces, even when there's occlusion between faces.

Real-Time Low-Bit-Rate H.263+ Video Codec

The goal of this project is to develop key video technology for videoconferencing. We propose novel techniques in order to fine-tune and enhance the coding performance for many applications.

Hand Tracking

Hand tracking can be used as a tool to determine hand gestures. We can apply the use of hand gestures to many daily applications of our lives, namely controlling a TV tuner or an entertainment system.

Sensor Fusion

In order to build a biometric system that is able to achieve desirable accuracy and operate efficiently, sensor fusion combining 2 or more different biometric approaches may be necessary.

Joint Source Channel Coding using Rate Shaping

To overcome the high error rate of the wireless channel as well as utilizing the bandwidth intelligently, we use rate shaping on coded video bitstream to deliver the video content effectively.

Traffic and Channel Modeling

We introduce a new stochastic process called the punctured  autoregressive (AR) process, and use it to model both the variable bit rate (VBR) video traffic and the wireless channel dynamics.