(A) study on the modulation parameter estimation, modulation type classification and self-recovery equalization of unknown cyclostationary signals

Abstract

Most of communication signals contain some periodicities such as carrier frequency, symbol rate, chip rate and so on. By these periodicities, the communication signals are in real cyclostationary processes rather than stationary processes. However, almost of conventional algorithms have been proposed on the assumption that the communication signals are stationary processes. In this dissertation works, we propose new algorithms for the modulation parameter estimation, modulation type classification and self-recovery equalization using the cycostationarities embedded in the unknown received signals. First, we review the characteristics of cyclostionarity and define some terminologies such as the cycle frequency, cyclic autocorrelation, cyclic spectrum and so on. Some formulae to get the cyclic spectrums are derived and the figures representing the magnitude of cyclic spectrum of the communication signals such as AM, PAM, QAM, BPSK, QPSK, SQPSK and MSK signals are shown. Next, we propose new estimation algorithms for the modulation parameters such as carrier frequency, carrier phase and symbol rate using the cyclic spectrum and the cyclic autocorrelation. For the carrier frequency estimation, the only one component of the cyclic autocorrelation, which is corresponding to the lag of zero, is used. This method is applied to a single sinusoid and BPSK signal. By the computer simulation, the performance of proposed method is verified and shown to be very robust against noise and run very fast. For the carrier phase estimation, we propose two methods using the cyclic spectrum and cyclic autocorrelation. The performance of proposed phase estimator is compared to the results of MLE and CRB. Through the performance evaluation, we can notice that the proposed method is comparable with MLE in terms of performance but the number of computations required in the proposed one is quite less than that of MLE. For the symbol rate estimation, we suggest two approaches, namely, on...