Signal detection on wireless CDMA downlink

On the downlink of a wireless code division multiple access (CDMA) system, single-user detectors have to be used because of the lack of the knowledge of the code, timing and power of other users. Moreover, since accurate channel modeling can be difficult, blind receivers and receivers based on training data rather than models are highly desirable. Developed from the universal classification theory, type-based receivers assume no a priori channel model, and are quite effective in direct-sequence spread spectrum single-user systems based on training data. We generalize the type-based receiver to multi-user systems. During the training stage, the exponential rates at which the error probabilities decay are maximized through optimal quantization. Our receiver operates on empirically observed sequences and rejects the unknown multiple access interference effectively in near-far situations.