Commercials will tell you to blame your cell phone carrier for problems with reception. They’re partly right, though location also plays a role — a Metro train might not be the best place for a heart-to-heart with a loved one.
But a seldom acknowledged source of bad reception comes from a basic challenge of signal processing. Cell phones convert data from two distinct channels of radio waves into information that can be used and understood. The same signal processing challenge applies for a wide range of devices used in communications, medicine and other fields.
For decades, engineers made various assumptions to simplify the data so they could design devices efficiently. They traded clarity for speed, reducing costs but also leaving us with garbled phone calls and reduced precision for some medical tests.
Now researchers at UMBC are receiving wide recognition for describing a mathematical technique that allows researchers to avoid this tradeoff.
“People have been making simplifying assumptions that don’t make good use of the data,” explains Tülay Adali, a professor of computer science and electrical engineering. “With the framework we developed, researchers don’t have to do that.”
The technique is not technically new. The researcher D.H. Brandwood described similar ideas in a 1983 paper. But Adali describes a “lightbulb” moment in 2006 when she was discussing that paper with Hualiang Li ’08, Ph.D., electrical engineering, who has since become a research assistant professor at UMBC.
The two researchers realized the technique could be applied broadly to process a range of data without having to simplify those assumptions.
“People had missed that one elegant, simple point,” Adali says.
Subsequent research at UMBC linked the technique to work published in the early 20th century by the Austrian mathematician Wilhelm Wirtinger. Adali and Li, along with Mike Novey ’09, Ph.D., computer science, and French researcher Jean-Francois Cardoso, described the technique and developed a framework for applying it to a range of problems in “Complex ICA Using Nonlinear Functions,” a paper published in IEEE Transactions on Signal Processing in 2008.
The four researchers were honored in May with the 2010 IEEE Signal Processing Society Best Paper Award, presented in Prague at the 2011 International Conference on Acoustics, Speech and Signal Processing. Given annually, the award recognizes up to six recent papers with significant impact on the field.
While cell phone reception problems frustrate many consumers on a daily basis, the same challenges present themselves in satellite communications, acoustics, radar, medical imaging and other fields. Adali’s research now focuses on an analysis of medical data from functional magnetic resonance imaging (fMRI) and other sources. Researchers often use fMRI to see what is happening in subjects’ brains while they are at rest or performing tasks such as driving or solving problems.
Applying Wirtinger’s techniques to analyze fMRI data, Adali and her colleagues have shown, on average, a 20 percent improvement in sensitivity detecting brain activity. The technique also does a better job highlighting which brain regions are involved in certain tasks, providing insights that can be used to more accurately diagnose mental disorders.
It’s a discovery that translates into shorter, more accurate tests and, for the broad field of signal processing, a more effective and efficient way to use data.
— Anthony Lane