HOS Based Goodness-of-Fit Testing Signal Detection

Detection performance of cognitive radios can be significantly improved by utilizing Higher Order Statistics (HOS) in the spectrum sensing process. The additional usage of goodness-of-fit tests on the inspected frequency spectrum can further improve the signal detection. This letter proposes a novel detection method that uses goodness-of-fit tests on the skewness and kurtosis of the FFT power spectrum (empirically derived) to evaluate the signal existence. Testbed evaluations prove that the method adapts to the specific hardware noise characteristics. The result is powerful detection performance, i.e. the capability of detecting various signals even 32dB below the noise floor.