Non-Data-Aided Phase and Frequency Offset Estimation for QAM Signals Using Batch Learning

Many algorithms for non-data-aided estimation of phase and frequency offset for signals with quadrature amplitude modulation are based on determining position and phase of the spectrum peak for nonlinearly transformed signal. In this paper the method is presented for obtaining required nonlinear weighting functions by batch learning from simulated training data. This approach allows to easily optimize algorithm for a wide range of real-world scenarios including those where optimum solution cannot be obtained in a closed form. Computer simulations confirmed applicability and promising results of the proposed method.