Teaching smart radio to sense wireless microphone signals
From left: Harpreet Dhillon, faculty advisor Mike Buehrer mike Benonis. Seated: Dinesh Datla, Jeong-O-Jeong.
While undergraduate competitions typically highlight design and teamwork, competitions at the graduate and research level can often be more about ideas. This is the case with the spring 2010 Qualcomm Cognitive Radio Competition, which Qualcomm designed to identify new ways of detecting wireless microphone systems. Virginia Tech was one of only 14 universities invited to participate.
The challenge stems from the recent switch from analog to digital TV signals and the FCC allowing cognitive radios to use TV white space. Cognitive radios may not interfere with existing communications and so must identify the signals of licensed users. The cognitive radios must have spectrum-sensing technology capable of detecting and avoiding ATSC and NTSC (digital and analog television signals), as well as wireless microphones.
Established techniques exist for sensing ATSC and NTSC, but detecting wireless microphones is more difficult. Wireless microphones can be found anywhere in the spectrum, have very low received signal strength and their transmitter format is not well defined. Moreover, multiple wireless microphones can coexist anywhere in the TV band, explained Harpreet Dhillon, a graduate student on the team. It was also required to detect the exact frequencies of all the wireless microphones present in the said frequency band.
“That was the biggest motivation for Qualcomm to float this competition. They wanted to get different ideas, because this problem is not so straightforward to solve. If it was, of course, those guys would have solved it already,” he said.
Dhillon and three other ECE graduate students competed for the $25,000 first prize. Michael Buehrer served as faculty advisor.
Dhillon had participated in design competitions as an undergraduate and found some differences in the experience. “Team spirit was absent in the initial phases of our efforts, because we are graduate students used to working independently on our own research. However, as our work progressed, we developed a unique spirit and sense of responsibility,” he recalled.
The team came together serendipitously, with each member contributing a different strength: Jeong O-Jeong acquired the literature, Dhillon developed the algorithm, Dinesh Datla characterized the noise, and Mike Benonis obtained real signals to create an extra data set for the team to test its algorithm.
The students developed a fairly robust solution. They determined a baseline noise correlation matrix from a data set with no wireless microphone signals. To see if a signal is present in a new environment, they compare the new correlation matrix with the baseline matrix using singular value decomposition. If their algorithm determines there is a signal, it calculates the center frequency from the measured auto correlation of the signal.
With no deliverables or hardware, the competition did not gather the teams together. The teams were given two data sets: a training set and a test set. When they were ready, they used their algorithm on the test set, which randomized the wireless microphone signals, and sent their results and algorithm to Qualcomm for evaluation. They expect to hear the results of the competition in April.
“It would have been nice to meet the other students,” Dhillon said. “When you delve so deeply into an issue, it’s always nice to meet other people and see how they dealt with the challenges. But in this case, people don’t always want to share their results. They may want to get IP or publish.”
Although different from his undergraduate experiences, Dhillon enjoyed the contest. “We had a good team and we appreciated how each person could contribute different skills,” he said. He also enjoyed a competition so related to his research. “In school, we develop the strong fundamental concepts. This was an eye-opener that helped me see the industrial applications of our work.”