| Communicating over wireless multipath channels is affected by InterSymbol Interference (ISI). Correct reception of the transmitted information symbols requires use of channel equalization techniques. Recent results in digital communications have shown that channel equalization can be achieved blindly, without transmitting a training sequence. In the first part of this dissertation, a new class of blind channel identification and equalization approaches for linear channels is proposed. New algorithms are developed to bypass drawbacks of existing techniques namely, sensitivity to the location of channel zeros, channel order overmodeling errors, and color of additive stationary noise. Performance of these algorithms is analyzed through comprehensive computer simulations and comparisons with the existing algorithms are presented. The asymptotic (large sample) performance of these algorithms is also investigated analytically. These channel equalization techniques are then extended to the more general problem of estimating the unknown ISI in the presence of carrier frequency offset and Doppler effects. The common feature of all these approaches is the diversity introduced at the transmitter by means of a digital periodically time-varying precoder. The second part of this dissertation proposes several techniques for blind equalization of nonlinear FIR Volterra and LTI-ZMNL-LTI communication channels. High Power Amplifiers (HPAs) are currently used in mobile radio and satellite communication channels. For an efficient utilization of the radio spectrum, high capacity, bandwidth efficient, and non-constant modulus Quadrature Amplitude Modulation (QAM) techniques are also employed. Due to the increased dynamic range of the input constellation, the HPA nonlinear characteristic and multipath, the overall communication link will be affected by linear and nonlinear signal distortions. Compensation of these undesirable distortions appears of high importance for designing high capacity/power efficient communication links. The proposed equalization techniques rely on the diversity introduced at the transmitter or the receiver by means of a precoder and/or by temporal/spatial oversampling of the received waveform, respectively. It is shown that multiple FIR Volterra channels can be blindly deconvolved using multichannel linear equalizers provided that the Volterra kernels satisfy a certain coprimeness condition. Finally, it is shown that equalization of LTI-ZMNL-LTI channels can be performed under quite general conditions without restrictions on the channel zeros or channel order over-estimation errors. |
