office:

Calit2 4th floor, the open space

email:

Please see the picture right below this column.

number:

(having trouble in reading?)

My research interests

My advisor is Prof. Bhaskar D. Rao. I am affiliated with Digital Signal Processing Lab in Dept. of ECE, Univ. of California, San Diego.

I am pretty interested in speech processing. My conjecture is that computer, or say machines, will ultimately be able to imitate human ability to hear, see, feel, infer, and, most important, shine creative sparks. Along this path, the following are of my special interests:

• Speech coding, recognition
• Multi-modality signal processing
• Intelligent systems

Another research turf that I am concerned about is source separation. Human have uncanny ability to separate mixed speech sources based on several kinds of cues. My question is: can machine achieve the same performance as what we do? Moreover, what if we extend our view to more general form of signals? Can signal separation problem be unified under a sound theory? Follow this path, I will focus on

• Source separation
• Compressed sensing (cf. website maintained by Rice Univ.)
• Inverse problem with sparse solutions
• Information theory

But of course, my current research is on Sparse Signal Processing. This is a domain that interests me (and bothers me quite enjoyably) quite a lot. I am taking a exciting ride on the powerful information theory and let it take me to the adventure. It's really amazing to see how concepts are essentially the same but with quite different interpretations in seemingly different domains. Be simple. Be what it's supposed to be.

Publications:

• Performance Limits of Matching Pursuit Algorithms, Yuzhe Jin, Bhaskar Rao, ISIT 2008, Toronto, Canada.

• Insights into the Stable Recovery of Sparse Solutions in Overcomplete Representations using Network Information Theory, Yuzhe Jin, Bhaskar D. Rao, ICASSP 2008, Las Vegas, USA.

• Spectral Estimation of Voiced Speech Using a Family of MVDR Estimates, Rajesh M Hegde, Yuzhe Jin, Bhaskar D. Rao, ICASSP 2007, Hawaii, USA.

Talks:

I have given several talks on different areas. The following are selected talks by me and my group mates.

• Capacity, Diversity and Multiplexing Gain for MIMO Channels

This is a talk for course ''Network Information Theory''. In this talk, we reviewed the fundamental tradeoff between diversity and multiplexing gains in MIMO channel performance evaluation.

• Sparse Solution to Linear Inverse Problems

In speech recognition course, I reviewed the paper on Sparse Solutions to Linear Inverse Problems using multiple measurement vectors (MMV). Matching pursuit type algorithms and Diversity minimization type algorithms are introduced.

• Robust Adaptive Beamforming with applications to Matched Field Processing

Working with my teammate Yong-Han, we studied several classic Robust Adaptive Beamforming algorithms with applications to Matched Field Processing.

• Adaptive Filtering with applications to Echo Cancellation

Joint project with Ting-Lan (Louis). We explored a classic method, PNLMS. We proposed our revision to the original method to improve the convergence performance.

Courses taken:

• Signal Processing

1. Statistical and adaptive signal processing
2. Parameter estimation
3. Speech recognition, coding

• Information Theory

1. Information theory
2. Network information theory
3. Source coding
4. Trellis-coded modulation

• Learning and AI

1. Statistical learning
2. Natural computing
3. Machine learning

• Misc.

1. Convex optimization

* Logo from Apple Pages 8.0