Optimal multiuser CDMA decoder using grouped coordinate ascent and the DWT

It is well known that direct implementation of the optimal multiuser receiver for an asynchronous code division multiple access (CDMA) communication system has an exponential complexity in the number of users. We present an iterative implementation of an optimal multiuser receiver which has at most cubic complexity per iteration in the number of users. The receiver performs joint maximum likelihood (ML) estimation of the amplitudes, phases, time delays, and symbol sequences of all the users. Similar, to the single user receiver we presented by Sharfer and Hero (see Proc. ICASSP, 1996), the multiuser receiver is based on a signal decomposition with respect to a compactly supported orthonormal wavelet basis. The algorithm uses grouped coordinate ascent updates to increase the likelihood function in each iteration. Simulation results of a two user system are presented showing that the receiver is near-far resistant, and attains the single user bit error probability bound.