Near-optimum decoding and performance bounds for multiple-antenna communication

Abstract

Multiple-antenna communication promises substantially enhanced data rate and diversity over rich-scattering fading channels. Coding for multiple-input multiple-output (MIMO) channels addresses three related problems: design of MIMO codes, MIMO decoding and performance analysis. In this dissertation, we look at selected topics in the latter two problems. In particular, we present near-optimum MIMO differential decoding algorithms, aiming at high decoding speed or improved performance in continuous fading, and derive tight performance bounds for space-time codes. The analysis of decoding algorithms and performance bounds herein may provide useful insights into code design and facilitate computer search for good codes. A number of MIMO differential decoders have been given in literature. However, most decoders are natural extensions of the traditional differential phase-shift keying (DPSK) techniques and do not really address some fundamental differences of the MIMO differential scheme. As a result, most decoders are limited to small constellation sizes and follow the linear prediction structure, which does not necessarily match the matrix signaling of MIMO. In this dissertation, we propose differential lattice decoding that is significantly faster for large constellation sizes. Moreover, noncoherent decoders wellmatched to the matrix signaling over continuously fading channels are presented. Proper approximations are made to obtain practical decoding complexity, which nonetheless guarantee near-optimum performance. The standard union bound is quite loose (in fact often divergent) for either coherent or noncoherent space-time codes. This phenomenon is especially pronounced in a quasistatic fading channel. We present tight performance bounds for space-time codes in the general framework of Gallager's bounding techniques. Closed-form upper bounds are proposed that are surprisingly tight in terms of the frame error probability. The proposed bounds only assume the knowledge of the weight spectrum, and can be evaluated very fast. In addition, we present novel methods of weight enumeration, which, in conjunction with the bounding techniques, give a complete treatment of performance bounds for space-time codes. The bounds are then extended to noncoherent MIMO schemes.