In this paper, maximum a posteriori (MAP) detection is applied to a direct-
sequence code-division multiple-access (I)S-CDMA) system jointly with ident
ification and estimation of time-selective Fading channels. By sampling the
outputs of the matched filter and combining antenna array elements, strong
and time-varying multiple-access interference (MAI) is characterized and s
uppressed instantaneously. The decision statistics for MAP detection are ob
tained from the conditional probability density function of prediction erro
r. The prediction is accomplished by approximating the fading channel with
a constrained nonlinear state model. Unknown parameters such as auto-regres
sive (AR) process coefficients, noise covariance matrices, and the antenna
array vector are estimated based on received sample vectors only. Also, dif
ferential modulation is applied to eliminate the need for pilot insertion.
Through computer simulations, near-optimum bit error rates (BERs) are found
.