USRP N210 demonstration of wideband sensing and blind hierarchical modulation classification

Blind modulation classification problem is particularly difficult when the exact frequency band of the signal is unknown since the modulation classifiers require accurate estimates of the signal parameters such as center frequency, bandwidth, and SNR. In this work, we demonstrate a hierarchical classification tree that filters and classifies a received signal as AM, FM, 4/16/64-QAM, 2/4/8-PAM, 4/8/16-PSK, DSSS, and FSK. Coarse estimates of signal parameters are obtained from energy detection and are refined using cyclostationary estimators. Cumulants and cyclostationarity are used to classify AM and FM while a reduced complexity Kuiper test is used for differentiating modulation level for QAM, PAM, and PSK. The effects of multipath are countered using a blind equalizer. The classifier is implemented in C++ using GNURadio libraries and is demonstrated using a USRP N210.