Liang Zou

Underdetermined Blind Source Separation: Theory, Methods and Applications

Why did you decide to pursue a graduate degree?

I wanted to do further research.

Why did you decide to study at UBC?

I am lucky to be a UBC'er. UBC provides us with the best environment for learning and conducting research that will serve the world's people. UBC is consistently recognised as a leading university, attracting scholars from across the globe.

What is it specifically, that your program offers, that attracted you?

A good supervisor.

What was the best surprise about UBC or life in Vancouver?

The view is amazing.

What aspect of your graduate program do you enjoy the most or are looking forward to with the greatest curiosity?

I'm most attracted to the flexible work style.

What do you see as your biggest challenge(s) in your future career?

Oral English is my biggest challenge.

What do you like to do for fun or relaxation?

I enjoy hiking in my spare time.

What advice do you have for new graduate students?

Enjoy your life here in Vancouver. You won't ever forget it.


Learn more about Liang's research

In my Ph.D thesis I focus on developing novel underdetermined blind source separation (BSS) methods and appling these methods in real-world applications.

  • To overcome limitations of currently available EMD-BSS based methods and recover the underlying source signals accurately, in my first project I propose a novel blind source separation framework. This framework combines noise-assisted multivariate empirical mode decomposition and multiset canonical correlation analysis. Upon applying the proposed method on the nano-sensor data, we found that the proposed framework can achieve better performance than other state-of-the-art approaches.
  • Existing BSS approaches are mainly designed for a single dataset BSS or the determined joint BSS problems (i.e., the number of sensors is greater than the number of sources for each dataset). To fill this gap, the main technical objective of my thesis focuses on developing underdetermined joint blind souce separation (UJBSS) approaches. In my second project, I exploit the second-order statistics of observations and introduce a novel UJBSS method which can extract the buried sources jointly from two datasets. Considering the dependence information between two datasets, the problem of jointly estimating the mixing matrices is tackled via canonical polyadic (CP) decomposition of a specialized tensor in which a set of spatial covariance matrices are stacked. Furthermore, the estimated mixing matrices are used to recover the sources from each dataset separately. I intend to further extend this idea to multiple datasets in a future paper.