Iterative Maximum Likelihood Decoding Ofgeneralized Spacetimeblock Codes

GENERALIZED SPACE-TIME BLOCK CODES A. Lee Swindlehurst Dept. of Elec. & Comp. Engineering, Brigham Young University, Provo, UT, USA, 84602 e-mail: [email protected] ABSTRACT An iterative algorithm for ML detection of a certain class of space-time codes is presented. Codes within this general framework employ redundant linear precoding, and include as special cases many recently proposed algorithms. A number of codes within this framework possess a special property that allows for decoupled ML symbol detection, which greatly reduces complexity. This property is exploited to derive an iterative minimizer of the ML criterion that alternates between estimating the transmitted data in one step, and the interference statistics and channel in the next. Training data is used to initialize the algorithm. INTRODUCTION The advantages of using multiple antennas at both the transmit and receive ends of a wireless communications link have recently been noted. A number of space-time codes have been proposed that exploit the potential for increased throughput and diversity that such systems offer. As shown in this paper, many of these codes can be unified within a general framework based on redundant linear precoding, where different linearly transformed versions of the same data sequence are broadcast from each transmit antenna. This framework is general enough to include the coding schemes described in [1]-[9] as special cases, among others. Codes within the proposed framework are referred to as Generalized Space-Time Block Codes (GSTBCs). In this paper, the special linear structure of GSTBCs is exploited for efficient Maximum Likelihood (ML) detection and estimation of the transmitted data sequence, the (flat-fading) channel coefficients, and the covariance matrix of the noise and interference. Sufficient training data is assumed to be available to obtain an initial estimate of the channel and the noise statistics, after which the algorithm iterates between estimating these parameters and detecting the symbols in the transmitted block of data. In [10], it was shown that for Alamouti-type codes [1, 5]), the symbol detection step can be decoupled into a series of scalar decisions, which greatly reduces algorithm complexity. In this paper, it is shown that this simplification results for many other codes within the GSTBC framework. GENERALIZED SPACE-TIME BLOCK CODES Assume a single-user transmit array with elements, a receive array with elements, and a flat-fading channel. Assuming symbol-rate samples are taken from the array, the following model results: