Blind modulation classification based on spectral correlation and its robustness to timing mismatch

Blind modulation classification which identifies the modulation class of a received signal with minimal a priori knowledge about the transmitter´s characteristics is crucial for both civil and military applications. Many modulation classification algorithms based on the spectral correlation function have already been proposed, but very few of them address practical aspects of its implementation and the impact of imperfect estimation of signal parameters. In particular, sampling at a non-integer multiple of the symbol rate has been shown to suppress the spectral correlation peaks of a modulated signal. In this paper, we propose a practical modulation classification architecture which blindly extracts the symbol rate from the received samples, recovers optimal samples through interpolation, and feeds them to a reduced-complexity modulation classifier based on cyclostationary feature detection. Using the proposed architecture, we evaluate the classification performance as a function of the number of samples and sampling mismatch, resulting from the symbol rate estimation error, under different SNRs for QAM, MSK, BPSK and AM signals.