Tuesday, April 24, 2007

"Help me if you can, I'm feeling down Louisiana close to New Orleans..."

Marty Baylor works on solving the "cocktail party problem" that keeps electronic devices from picking out individual voices from a group. Her research at CU could change communications systems, her adviser says.

Baylor studies mixed signals, unscrambling them with a laser system.

Leaning over a delicate setup on a laboratory table, long braids pulled back in a practical headband, Baylor points out refractive crystals, modulators, mirrors and beam splitters.

She describes the "cocktail party problem" that she and others are trying to solve: Human ears and brains are great at picking out single voices from a group, but getting electronic devices to do the same task has proved vexing.

An easy and quick solution could improve hearing aids, cellphones, and help intelligence agents eavesdrop on enemy communications.

It also would enable scientists to sort out signals from a set of robots sent to a distant planet.

For nonphysicists, Baylor and her colleagues have developed a demonstration - she creates a mix of the Beatles' "Help," and Chuck Berry's "Johnny B. Goode" on a computer.

At first, it's cacophony.

Then, within seconds, Berry's guitar fades, and only "Help" remains.

"As far as I know, we're the only people in the world who solve the problem this way," Baylor said. (more)