Cyclic frequency distinction and its application in modulation parameter estimation: An adaptive approach

Cyclic frequency detection of cyclostationary signals and its application in modulation parameters estimation is the aim of this paper. A simple frequency shifted (FRESH) model is structured. The normalized least mean square (NLMS) algorithm is engaged to estimate the unknown weights of the considered model. The norm of the parameter estimate is related to the model accuracy. A high (low) value indicates that the FRESH structure is based on an accurate (inaccurate) cyclic frequency and the weight estimate of the NLMS algorithm converges to a nonzero (zero) vector. Accordingly for two given values which only one of them is the true cyclic frequency of the received signal, two FRESH structures are considered and simultaneously two NLMS algorithms are executed. If the norm of the weight estimate of the first algorithm exceeds the resulted norm of the second one, according to the proposed method, the first given value is the true cyclic frequency and vice versa. The method is used to estimate the modulation parameters of three different problems: chip rate estimation of a spread spectrum signal, carrier frequency and bit rate estimation of BPSK and QPSK signals, and symbol rate estimation of OFDM signals.