Lattice-based subspace decomposition for DS-CDMA detection

Lattice filtering is applied to the MMSE (minimum mean squared error) DS-CDMA receiver. The lattice performs Cholesky decomposition on the sampled input covariance matrix. This identifies information about signal and noise subspaces, which can be exploited to decrease the complexity of the MMSE receiver using LS (least squares) adaptive algorithms. Two reduced complexity techniques are considered: (1) filter truncation and (2) linear combining of multiple truncated subfilters. In both cases, the dimension of the (sub)filters must be larger than that of the user subspace. Numerical results presented compare the performance of these techniques using LMS and LS algorithms.