Techniques de transmission et de modélisation de canal pour les systèmes de communications multi-antennes

This thesis deals with different aspects of multiple-antenna (MIMO) wireless communications. In a first part, we introduce Space-Time-Frequency Spreading (STFS), a space-time code for Orthogonal Frequency Division Multiplexing (OFDM) systems that maximally exploits the channel space and frequency diversity, as well as time diversity. For low spectral efficiencies (e.g. BPSK), Viterbi-based decoding of STFS associated with iterative interference cancellation provides the same performance as BICM without requiring soft decoding. In a second part, the importance of having channel state knowledge at the transmitter is discussed. A channel reciprocity model is introduced for the case of Time-Division Duplex (TDD) systems, which models the impairments of the radio-frequency components with linear filters. After a collaborative training phase (relative calibration), this model enables the transmitter to infer the downlink channel impulse response from the uplink channel estimates, thus lifting the requirement for continuous feedback. The frequency-selective reciprocity model was experimentally validated for SISO channels. Finally, the problem of modeling the temporal evolution of MIMO frequency-selective channels is addressed. A pathwise model is introduced, and we propose to use a blind method to decompose the time-varying channel realizations into, for each path, a set of constants representing the physical characteristics of the environment, and a time-varying, structured process (such as a Doppler series) that can be easily tracked or predicted. The performance of this method is evaluated by simulations, using both synthetic and experimental data.