Doctor of Philosophy

dissertationThis dissertation deals with blind modulation identification of quadrature amplitude modulations (QAM) and phase-shift keying (PSK) signals in dual-polarized channels in digital communication systems. The problems addressed in this dissertation are as follows: First, blind modulation identification of QAM and PSK signals in single noisy channels and multipath channels are explored. Second, methods for blind separation of two information streams in a dual-polarized channel and identification of the modulation types of the two information streams are developed. A likelihood-based blind modulation identification for QAM and PSK signals in a single channel with additive white Gaussian noise (AWGN) is developed first. This algorithm selects the modulation type that maximizes a log-likelihood function based on the known probability distribution associated with the phase or amplitude of the received signals for the candidate modulation types. The approach of this paper does not need prior knowledge of carrier frequency or baud rate. Comparisons of theory and simulation demonstrate good agreement in the probability of successful modulation identification under different signal-to-noise ratios (SNRs). Simulation results show that for the signals in AWGN channels containing 10000 symbols and 20 samples per symbol, the system can identify BPSK, QPSK, 8PSK and QAMs of order 16, 32, 64, 128 and 256 with better than 99% accuracy at 4 dB SNR. Under the same condition, the simulation results indicate the two competing methods available in the literature can only reach at most 85% accuracy even at 20 dB SNR for all the modulation types. The simulation results also suggest that when the symbol length decreases, the system needs higher SNRs in order to get accurate identification results. Simulations using different noisy environments indicate that the algorithm is robust to variations of noise environments from the models assumed for derivation of the algorithm. In addition, the combination of a constant modulus amplitude (CMA) equalizer and the likelihood-based modulation identification algorithm is able to identify the QAM signals in multipath channels in a wide range of SNRs. When compared with the results for the signals in AWGN channels, the combination of the CMA equalizer and the likelihood-based modulation identification algorithm needs higher SNRs and longer signal lengths in order to obtain accurate identification results. The second contribution of this dissertation is a new method for blindly identifying PSK and QAM signals in dual-polarized channels. The system combines a likelihood-based adaptive blind source separation (BSS) method and the likelihood-based blind modulation identification method. The BSS algorithm is based on the likelihood functions of the amplitude of the transmitted signals. This system tracks the time-varying polarization coefficients and recovers the input signals to the two channels. The simulation results presented in this paper demonstrate that the likelihood-based adaptive BSS method is able to recover the source signals of different modulation types for a wide range of input SNRs. Comparisons with a natural gradient-based BSS algorithm indicate that the likelihood-based method results in smaller symbol error rates. When a modulation identification algorithm is applied to the separated signals, the overall system is able to identify different PSK and QAM signals with high accuracy at sufficiently high SNRs. For example, with 20,000 symbols, the system identified BPSK and 16-QAM signals with better than 99% accuracy when the input SNR was 8dB and the polarization coefficients rotated with a rate of 1.3 ms. Higher SNRs are needed to obtain similar levels of accuracy when the polarization changes faster or when the number of input symbols is shorter. When compared with the identification results for signals in AWGN channels, the system needs higher SNRs and longer signal length to obtain accurate results for signals in dual-polarized channels