BIMA: BLIND ITERATIVE MIMO ALGORITHM T. Dahl, N. Christophersen, D. Gesbert. Department of Informatics, University of Oslo, P.O. Box 1080, N-0316 Blindern, Norway e-mail: tobias@ifi.uio.no ABSTRACT Identification of the channel matrix is of main concern in wireless MIMO (Multiple Input Mul- tiple Output) systems. Here, we present an SVD- based approach for blind identification of the main independent parallel channels. The right and left singular vectors are estimated directly (no channel matrix estimation is necessary) and continuously updated during normal transmis- sion. The approach is related to the iterative Power Method [8], as well as the time reversal approach ([4]). 1. INTRODUCTION Wireless MIMO systems are capable of deliver- ing large increases in capacity through utiliza- tion of parallel communication channels [5], [6], [12]. Fora N (receive) × M (transmit) channel matrix H of rank K 0 ≤ min(N,M ), the parallel channels are naturally realized through the Singular Value Decomposition (SVD) H = USV H , when the channel matrix is known both at the transmitter and the receiver side. S is the diagonal matrix of singular values σ 1 ≥ σ 2 , ··· ,σ K0 > 0, and U =[u 1 ,..., u K0 ] ∈ C N,K0 (1) V =[v 1 ,..., v K0 ] ∈ C M,K0 (2) are unitary matrices whose columns can be used as receive and transmit vectors {u i } and {v i }, respectively. One can select a number K (K ≤ K 0 ) of transmit/receive vectors to use for communication. Under stationary conditions, one may try to determine H experimentally and subsequently perform the SVD as in the sonar application [10]. For time-varying systems, most studies have assumed that H is unknown at the transmitter and known - through training data - at the receiver. However, first, this implies overhead, and second, the use of channel knowledge on the receiver only leads to less efficient use of the MIMO system. The transmit array diversity gain is not realized, and one is unable to transmit on the top singular vectors, those giving maximum performance/complexity tradeoff. In the method presented, two-way transmis- sion of data allows the two parties to estim- ate a selected set of left and right singular vec- tors, without explicit knowledge of H. Unlike other previous methods for blind MIMO estim- ation (for example [13] and references therein), which rely on a statistical based estimation of the channel matrix, our technique estimates the eigen-structure of the MIMO channel directly, without need of an actual SVD. The key advant- age of this technique is that it exploits trans- mission of regular symbol data to acquire an up- date of the singular vectors. 2. METHODS Assume a flat fading MIMO channel H exhibiting reciprocity. The uplink and downlink channels are the same, as in TDD (Time Division Duplex) systems. Without noise, transmission (s) and receiving (r) for two parties X and Y (for instance, X =base III - 2365 0-7803-7402-9/02/$17.00 ©2002 IEEE