Date Added: Dec 2009
The authors consider the problem of joint Non-Data-Aided (NDA) estimation of carrier frequency and phase offsets, and the signal and noise powers, from the baud-rate samples of a linearly modulated signal in Gaussian noise. The Cramer-Rao Lower Bound (CRLB) is obtained, showing that, for quadrature-symmetric constellations, the carrier parameters are decoupled from signal and noise powers. A joint NDA Maximum-Likelihood (ML) estimator is developed, based on the application of the Expectation-Maximization (EM) algorithm. Its performance is close to the CRLB for a wide SNR range.