Systems and methods for cognitive radio spectrum sensing of a signal are disclosed herein. On exemplary method comprises applying a pre-defined cyclostationary feature to detect the presence of the signal; detecting the signal; detecting a spectrum associated with the signal; sampling randomly the detected signal from its cyclic frequency domain; and applying a compressive sensing algorithm to classify the signal based on the cyclostationary feature. The signal can be sparse in time, space, frequency, or code domains. Thereby, the systems and methods described in the present disclosure involve exploiting compressive sensing in a specific sparse domain (i.e., cyclic domain) and also utilize a cyclostationary feature based compressive spectrum sensing scheme to perform spectrum analysis.