| In conventional wireless communication systems, training sequences are transmitted periodically in order to track the time-varying channel environments. This is neither power efficient, nor spectrally efficient. As an effort to improve the spectral efficiency by reducing the overhead, this dissertation is focused on blind channel estimation and signal detection for direct sequence spread spectrum systems.; In literature, if the spreading sequences are periodic and repeat with every information symbol, the system is referred to as short-code CDMA, and if the spreading sequences are aperiodic or essentially pseudo-random, it is known as long-code CDMA. The time-varying nature of long-code CDMA significantly complicates the development of blind detectors, as needed statistics can no longer be obtained through time averaging of the observed data. For this reason, research on blind multiuser detection has largely been limited to short-code CDMA. On the other hand, long-code is widely used in virtually all operational and commercially proposed CDMA systems due to its inherent security features and performance stability in frequency fading environment.; In this dissertation, statistics based algorithms are developed for both short-code and long-code DS-CDMA systems, with emphasis on the long-code systems. The major contributions of the dissertation can be briefly summarized as: (i) Blind detectors based on the code-constrained super-exponential algorithm have been developed for multi-rate short-code CDMA systems, while only the spreading code of the desired user is assumed to be known. (ii) For long-code CDMA systems, considering the time variant nature of the spreading code, the chip-rate scrambled signal is taken as the system input, and the long-code CDMA is characterized using a time invariant model. (iii) In downlink long-code systems, after chip-level equalization, the descrambled signal is treated as the received signal of a short-code CDMA system, and super-exponential algorithm is applied to recover the information symbols. (iv) For uplink, two-stage approaches have been developed in this research. In the first stage, multi-step linear-prediction-based methods are developed to eliminate the inter-symbol interference. In the second stage, if the spreading codes are of nonconstant modulus, blind channel estimation is performed by exploiting the second-order statistics; if the spreading codes are of constant modulus, higher-order statistics based algorithms need to be applied to estimate the channels. (v) To further improve the transmission quality without increasing the transmission power or bandwidth, time-reversal space-time block coding (TR-STBC) scheme is applied at the base-station side in downlink CDMA, and blind signal detection algorithm based on the principal component analysis has been developed.; Throughout this research, computer simulations are carried out to illustrate the proposed approaches. |
