Blind channel estimation and multiuser detection for CDMA communications

For high performance CDMA communications, multiuser detection is required to suppress the multiple access interference and fully exploit the system capacity. This requires the knowledge of users' signature waveforms which, in wideband CDMA, depend on channel parameters such as multipath coefficients and residual carrier offsets. These parameters need to be estimated in practice. This thesis studies channel estimation and multiuser reception for CDMA communications without the use of a training sequence. We exploit the structure and redundancy inherent in CDMA signals to develop several blind channel acquisition and multiuser detection schemes. More specifically, our contributions include the following: (1) A novel subspace algorithm to carry out joint carrier and multipath estimation blindly: The convolutional channel effect and multiplicative carrier drift are decoupled with a matrix polynomial operation and unknown parameters are estimated in closed-form. (2) A maximum likelihood channel estimator (MLE) for CDMA systems using aperiodic spreading codes: The multidimensional optimization problem is solved using an iterative algorithm based on the alternating maximization principle. (3) A direct demodulation method for low complexity uplink reception: The algorithm is an extension of the constrained MOE detectors by taking into account the channel effects. (4) A blind channel equalizer for downlink CDMA communications: An algorithm exploits the fact that all signals go through a common channel. The proposed equalizer is easy to be implemented in real-time and has performance close to that of an ideal ZF receiver.