Sub-Nyquist sampling and machine learning based online automatic modulation classifier for multi-carrier waveform

Reconfigurable terminals capable of adapting their transmission parameters such as modulation scheme, data and coding rates are desired to enhance the quality of service along with energy and spectral efficiency. However, spectrum bandwidth and time delay constraints limit the explicit sharing of these parameters and hence, they need to be estimated blindly. Though significant work has been done in blind parameter estimation, the task becomes non-trivial when reconfigurable terminals employ sub-Nyquist sampling (SNS) for wideband signal digitization. The proposed work presents the SNS and blind reconstruction based automatic modulation classifier (AMC) to blindly identify the modulation scheme of wideband multi-carrier signal, e.g. OFDM waveform. Simulation results along with experimental results on the proposed USRP testbed show that the classification accuracy of SNS based AMC approaches to that of Nyquist sampling based AMC with increase in signal to noise ratio given that the received signal is sufficiently sparse in the frequency.