Frequency domain iterative pulse shape estimation based on second-order statistics

This work addresses the blind identification problem of the received pulse shape in linear modulations. In particular, the unconditional maximum likelihood approach is adopted, with special emphasis on the application to working conditions with low signal-to-noise ratio. The identification procedure is based on the exploitation of the second-order cyclostationary property of the received signal, and for this purpose, a cost function for the problem at hand is formulated by extending the signal model to the frequency domain. In this sense, an iterative algorithm is proposed which results in a fast convergence and a robust performance in the presence of noise.