Multiuser detection for UTRA-TDD based on polynomial expansion

In this paper the use of the polynomial expansion (PE) detector as a multiuser detector (MUD) for UTRA-TDD is discussed. The PE detector can iteratively approximate either the decorrelating or the MMSE (minimum mean-squared error) MUD. This is possible without the computationally intense matrix inversion operation (cubic complexity in the number of users: O(K/sup 3/)), which is needed for the decorrelating and the MMSE detector. In this paper it is shown that the operations for the PE detector can be arranged such that they only have a complexity of O(K/sup 2/), in all processing levels. New simulation results for an UTRA-TDD simulation environment show that the detector achieves a performance (in terms of BER) close to that of the MUD which it approximates (decorrelating or MMSE). Additionally, it is shown how the number of iterations can be reduced significantly, while achieving the same BER performance.