Spectrum window based signal detection at low SNR

Signal detection is of interest in spectrum awareness that is useful for future cognitive military radios for their spectrum usage decisions. Detection is often based on energy on a narrow bandwidth or frequency bin. For wideband signals covering several frequency bins a spectrum matching detector has been proved to be more sensitive. This requires correlation with several a priori known candidate spectra making it more complicated than the single bin detector. Herein, a simplified version of spectrum matching detector using a rectangular reference spectrum is investigated. It is shown that the proposed simplified version outperforms the single bin version. It can detect wideband signals even below noise level. Furthermore, it was observed that there are several issues that should be solved before the detector could be used in practice. One is that many rectangular window sizes can detect the same signal and, as a consequence, one should find a method to decide which one is the correct one.