Approximate and exact ML detectors for CDMA and MIMO systems: a tree detection approach

Abstract

This paper deals with Direct-Sequence Code Division Multiple Access (DS-CDMA) transmissions over mobile radio channels. Different detection techniques have been proposed in the past years, among which approximate ML detectors. In this paper we propose an exact ML detector with a low complexity. We show that a QR factorization of the matrix of users’ signatures is appropriate: the upper triangular form of the R matrix makes it possible to state the ML detection in terms of a shortest path detection in a tree diagram. Different algorithms based on the tree diagram are compared: the stack algorithm (exact ML), and the feedback decoding algorithm (approximate ML). The numerical complexity of the proposed techniques is studied in detail; in particular, the low complexity of the stack algorithm at high SNR is pointed out. Simulations show that the performance of the stack algorithm in terms of Bit Error Rate (BER) are very close to the single user bound. Furthermore, we point out the fact that the same kind of approach can be used to perform ML detection in Multiple Input Multiple Output (MIMO) systems.