Universal classification for CDMA communications: single-user receivers and multi-user receivers

Universal classification refers to the empirical detection problem based on training sequences without assuming any a priori models, and the theory finds important applications in signal detection in the presence of uncertainties. Because the channel characteristics of a code division multiple access wireless system are difficult to model, we develop type-based receivers that operate in a data-driven fashion based on universal classification theory. On the downlink, we focus on the type-based single-user receiver that is capable of multiple access interference suppression with a reasonable amount of training data. We extend universal classification theory to interleaved stationary Markov sources, and propose the type-based multi-user receiver for the uplink. The error probability of type-based receivers decay exponentially asymptotically regardless of the nature of the channel noise