Aggie, Ingeniero January 2011

NMSU engineering professors use state-of-the-art equipment in research

As New Mexico State University’s research becomes more dynamic, complex and innovative, NMSU professors and student researchers are working to collect and process more data, more efficiently, and faster.

Charles Creusere and Phillip De Leon, both electrical engineering professors at New Mexico State University, are utilizing state-of-the-art graphics processing unit computing processors to enhance their research.

GPU processors are specialized microprocessors that accelerate two-dimensional and three-dimensional graphics rendering within devices such as personal computers, and game consoles but can also be used for general-purpose computing and simulation. Using state-of-the-art GPUs, research that 20 years ago would have taken a supercomputer, can now be performed using a processor able to fit within a regular desktop computer, cutting processing time the equivalent of a month into a couple of days.

The new GPU processors can boost a system’s performance 10-100 times as fast as a regular PC. Whereas a PC’s CPU may have two or four cores, the NVIDIA GPUs being used have as many as 448 cores, enabling complex simulation.

Creusere, who has been researching humans’ perception of quality changes in audio, utilizes the NVIDIA Tesla to collect and sift through the immense amounts of data collected through electroencephalography (EEG) brainwave testing. EEG testing measures the brain waves being produced by a subject’s brain through128 electrodes that are placed on the subject’s head. Data is collected from each channel under several different conditions and then has to be filtered.

“We don’t just have one test subject, we have 20 test subjects and each one of them has listened to the same audio with the variations multiple times so we have to cross correlate all of this. It’s like sifting a needle through a haystack to figure out which of these variations are consistent across different audio frequencies,” Creusere said.

Creusere looks to determine how quality of audio is perceived by a human listener, by measuring directly from the brainwaves of that listener by applying various forms of temporal and spatial digital processing in which they localize and identify brain signals that correlate with the perceived quality of the audio changes.

“We will use these models to improve the quality of audio that people listen to which is being transmitted digitally, such as high-definition radio, satellite radio, Internet radio, etc.,” Creusere said.

Creusere, director of the Center for Telemetry and Telemetering at NMSU, in partnership with Jim Kroger, associate psychology professor, have received a $38,762 Interdisciplinary Research Grand Award from NMSU for this research project.

Utilizing a similar GPU, De Leon, director of the Advanced Speech and Audio Processing Laboratory at NMSU, plans to apply high-performance computing using GPUs to digital signal processing in several areas of research. His research is supported by a $140,000 grant from the U.S. Army.

De Leon, in collaboration with electrical engineering graduate student Steven Sandoval and Research Assistant Professor Laura Boucheron, is developing new methods to code speech signals with as few bits as possible while still being intelligible.

“We use the GPU to perform many of the complex signal processing operations at both the encoder and decoder stages in parallel,” De Leon said. “Some of our signal processing routines are running 40 times faster than before.”

De Leon, graduate student Richard Gutierrez and undergraduate Greg Hinojos are also developing new methods for recognizing and identifying persons from live audio streams. The GPU will help train the system on voice signals much more rapidly, reducing the training time by a factor of 10.

And Don’t Miss These Stories