The effectiveness of Savitzky-Golay smoothing method for spectrum sensing in cognitive radios

The paper proposes a new smoothing method to reduce noise contribution in frequency spectra, whose analysis is accomplished in order to perform spectrum sensing task for Cognitive Radio applications. The smoothing phase is a fundamental step in frequency analysis, since noise often corrupts user signals, by preventing them to be detected and, consequently, either protection or demodulation become impracticable. Such a need is usually satisfied through the use of moving average filters. These kinds of filters are affected by some problems which make the sensing stage not accurate and scarcely reliable. The authors propose the employment of Savitzky-Golay filters, which are already widely used in biomedical image analysis, but they are almost absent in Cognitive Radio field, to our knowledge. The goodness of such an approach is proved by two different figures of merit, testing the filtering abilities and the sensing performance improvement, thanks to the previous smoothing stage. A comparison is finally proposed with the standard linear moving average method.