Amplitude normalization in blind modulation classification

The classification of digital modulation schemes plays an important role in communication intelligence (COMINT) and other related applications. The existing algorithms for modulation classification consider a semi-blind scenario, where certain signal parameters are assumed to be known. The pre-processing accuracy of signal parameters like the symbol rate, the center frequency, the carrier phase and the signal amplitude etc. has direct implication on classification. Here we address the case of model mismatch due to the amplitude uncertainty in maximum likelihood (ML) classification and propose a new approach to mitigate the situation. The method is based on the normalization of received signal amplitude using fuzzy clustering algorithm. Simulation results are presented to show the robustness of the algorithm for blind scenario. Concluding remarks are made with the scope for future work.