2/15/2013 Megan Kelly
Written by Megan Kelly
ECE graduate students Sreekanth Annapureddy, Sreeram Kannan, Adnan Raja and Jayakrishnan (Jay) Unnikrishnan won the 2010 Qualcomm Cognitive Radio Contest for developing algorithms and software to make radio spectrum use more efficient. The prize carries a $25,000 monetary award. In addition, the group’s research will be included in a book on cognitive dynamic systems.
Eleven teams from seven universities competed in the contest. CSL Researcher Venu Veeravalli, ECE, advised the winning team.
"The students were required to work on the project independently without any faculty input. I simply served as Qualcomm's liaison with the students. My only contribution to the project, if any, was that I taught three of the four students the techniques they needed for the project in my Detection and Estimation (ECE 561) class," Veeravalli said.
"I am very proud of their achievement," Veeravalli added.
Radio spectrum space is scarce and very expensive to license, but only one-fifth of the spectrum is normally in use at a given time, Kannan said.
For example, Krannert owns a certain amount of the spectrum for their wireless microphones,” Kannan said. “However, Krannert only uses wireless microphones during events. When they aren’t using the bandwidth, there’s a proposed regulation that will allow others to transmit signals on that spectrum.”
The devices that detect vacancies in the spectrum are called cognitive radios. The trick to using this open bandwidth is determining when it’s free, which was the objective of the Qualcomm Cognitive Radio Contest.
The Qualcomm contest requires students to develop an algorithm and software that detects if a wireless microphone signal is present within a spectrum, so the cognitive devices can avoid causing interference to it. The wireless microphone needs to be protected from interference as it holds the primary license for that spectrum.
However, finding the wireless microphone signals proved harder than they imagined. Raja said that Qualcomm provided the contestants with “training data,” which specified the frequencies where microphone signals were located. Contestants were expected to use this information to learn models and develop algorithms to classify the signals.
“The spectrum they provided was six megahertz wide, but the wireless microphone signal is very narrow, less than 200 kilohertz wide, and it can be anywhere in the spectrum, which makes it hard,” Unnikrishnan said.
“It’s like finding a needle in a haystack,” Raja added. “It’s so narrow and it can be anywhere in the spectrum.”
Kannan said the key problem was learning to distinguish wireless microphone signals from unintended narrowband interference caused by electrical devices in the vicinity.
"We solved this by identifying features that characterize narrowband interference," Kannan said.
The algorithm was judged based on its performance, novelty and implementation complexity.
“We needed to keep in mind its practical applications, which was hard because we’re used to working with the theoretical,” Kannan said.
The group persisted and developed a strong, unique algorithm that classified signals with good confidence levels, which contributed to their win. If implemented into Qualcomm applications, their algorithm could help target open bandwidth and channel it into use.
“One potential application of cognitive radios is to bring high speed Internet to rural areas where it’s not presently available,” Raja said.
The winning students look forward to the future of their research.
“I think it’s very impactful in this sense. These kinds of competitions help us connect theory to practical applications and even guide futuristic research,” Kannan said. “It makes it all exciting.”