Linear and nonlinear chip-rate minimum mean-squared-error multiuser CDMA detection

Abstract

A linear Kalman filter detector for code-division multiple access proposed earlier in the literature is extended to a structure that can handle arbitrary detection delays, through the mechanism of state augmentation. Because pre-detection RAKE combining is used in the detector, it is optimal for multipath channels, unlike the previous structure that performed post-detection combining. We also derive nonlinear Kalman detectors, which approximate the highly complex nonlinear minimum mean-squared-error detector, using the concept of additional observations. Both linear and nonlinear detectors require processing at one or more times the chip rate, and knowledge of the spreading codes of interfering users. They have the advantage over many other multiuser detection algorithms of not requiring the spreading codes to be periodic at the symbol rate, or matrix inversion. In addition, two of the detectors are able to generate and update a posteriori probabilities of the transmitted symbols, making them interesting for iterative multiuser detection.