Independent Component Analysis-Based Source Separation with Noise Cancelling for Cyclostationary Detection in Cognitive Radios

Cyclostationary detection was shown to be a promising technique for signal detection and identification in cognitive radios (CRs). In the presence of multiple transmitted signals, however, the corresponding cyclic frequencies become superposed in the cyclic profile, making it impractical to identify the cyclic features of each signal. To address this issue, we propose a spectrum sensing technique based on blind source separation (BSS) which separates the active sources using independent component analysis (ICA) before estimating the cyclic frequencies. The source separation performance is enhanced by applying an adaptive noise cancelling (ANC) filter to the received signals. Cyclic profiles with improved estimation accuracy are then obtained by computing cyclic profiles of each source signal separately. We evaluate the source separation performance of the combined ANC-ICA technique and show its superior performance, compared to other alternative source separation techniques.