Antenna diversity-assisted adaptive wireless multiuser OFDM systems

Single-user decision-directed channel estimation for OFDM is known to deliver potentially more accurate channel transfer factor estimates than pilot-assisted schemes, when benefiting from error-free symbol decisions. However, the estimates' accuracy is degraded in the context of rapidly fluctuating channels, since the channel transfer factor estimates produced during the previous OFDM symbol period are employed for the frequency-domain equalization of the most recently received OFDM symbol. Hence, the employment of Wiener prediction filtering was shown to be an effective countermeasure for mitigating the effects of channel transfer function variations imposed by higher Doppler frequencies. Two techniques were compared against each other, namely a scheme, which was insensitive to the shape of the CIR encountered and an adaptive prediction filtering. Furthermore, the joint effects of employing decision-directed channel prediction and adaptive modulation were demonstrated. Based on the philosophy of a decision-directed channel estimator designed for single-user scenarios, a parallel interference cancellation assisted channel estimator was proposed for multi-user scenarios, or more generally for OFDM systems employing multiple transmit antennas. This was motivated by the observation that the utilization of the most prominent subspace-based least-squares channel estimator is restricted to scenarios, where the number of users supported is lower than or equal to the number of OFDM subcarriers normalized to the number of significant CIR-related taps to be estimated. An iterative procedure was proposed for the off-line optimization of the estimator's coefficients. Alternatively, an adaptive approach based on the recursive least-squares (RLS) algorithm was proposed for updating the channel estimator coefficients on an OFDM symbol-by-symbol basis. Furthermore, a suite of detection techniques to be employed in a multi-user SDMA-OFDM scenario was compared against each other. Specifically, least-squares detection (LS), minimum mean-square error (MMSE) detection, successive interference cancellation (SIC), parallel interference cancellation (PIC) and maximum likelihood detection (ML) were studied. Detailed investigations were conducted with respect to the effects of error-propagation potentially occurring across the different SIC detection stages. Various strategies designed for improving the standard SIC detector's performance based on tracking multiple symbol decisions from each SIC detection node were compared against each other. An improved soft-bit metric, which takes into account the effects of error propagation was proposed for a system employing both SIC detection and turbo-decoding. Investigations conducted for a system employing PIC-detection and turbo-decoding demonstrated that potentially the same performance as that of the SIC-detection assisted turbo-decoded system can be achieved, although at a lower complexity. Finally, the performance of the low-complexity MMSE- or PIC detection aided systems was further improved by employing adaptive modulation or Walsh-Hadamard Transform based spreading, with the aim of exploiting the wideband channel's diversity potential in the detection process.