Unimodular-Upper polynomial matrix decomposition for MIMO spatial multiplexing

We present a simple algorithm to compute the factors of a Unimodular-Upper (UU) polynomial matrix decomposition. The algorithm relies on the classical LU factorization and the inverse of the unimodular factor is also provided. Such decomposition is useful for spatial multiplexing in MIMO channel transmission system since it enables to reduce the MIMO channel matrix into independent SISO channels by a pre- and post-filtering. Unlike the classical QR-based polynomial matrix Singular Values Decomposition (QR-PMSVD), the proposed UU method allows to completely cancel the co-channel interference (CCI). Moreover, most of the resulting independent SISO channels are likely to be reduced to simple additive noise channels, i.e. with no InterSymbol Interference. However, the noise is coloured and possibly enhanced due to the non unitary property of the corresponding post filter. The complexity and sum rate capacity performance of the proposed method are studied and compared with QR-PMSVD.