Seminar on covariance-based spectrum sensing

29.09.2014 20:05

Here are the slides from my seminar on a practical test of covariance-based spectrum sensing methods. It was presented behind closed doors and was worth five science collaboration points.

Covariance-based spectrum sensing methods in practice title slide

The basic idea behind spectrum sensing is for a sensor to detect which frequency channels are occupied and which are vacant. This requires detecting very weak transmissions, typically below the noise floor of the receiver. For practical reasons, you want such a sensor to be small, cheap, robust and capable of detecting a wide range of signals.

Covariance-based and eigenvalue-based detectors are a relatively recent development in this field. Simulations show that they are capable of detecting a wide range of realistic transmissions, are immune to noise power changes and can detect signals at relatively low signal-to-noise ratios. They are also interesting because they are not hard to implement on low-cost hardware with limited capabilities.

Over the summer I performed several table-top experiments with a RF signal generator and a few radio receivers. I implemented a few of the methods I found in various published papers and checked how well they perform in practice. I was also interested in what kind of characteristics are important when designing a receiver specifically for such an use case - when using a receiver for sensing, noise properties you mostly don't care about for data reception start to get important.

This work more or less builds upon my earlier seminar on spectrum sensing methods and my work on a new UHF receiver for VESNA. In fact, I have performed similar experiments with the Ettus Research USRP specifically to see how well my receiver would work with such methods before finalizing the design. Since I now finally have a few precious prototypes of SNE-ESHTER on my desk, I was able to actually check its performance. While I don't have conclusive results yet, these latest tests do hint that it does well compared to the bog-standard USRP.

A paper describing the details is yet to be published, so unfortunately I'm told it is under an embargo (I'm happy to share details in person, if anyone is interested though). But the actual code, measurements and a few IPython notebooks with analysis of the measurements are already on GitHub. Feel free to try and replicate my results in your own lab.

Posted by Tomaž | Categories: Life

Add a new comment


(No HTML tags allowed. Separate paragraphs with a blank line.)