This thesis studies the performance of the multi-source relay cooperative network where the linear algebraic Network Coding is applied at the relay. We work in the wireless sensor network context. We consider a cooperative coding scheme based on turbo product codes where a number of sensors transmit to a same destination with the help of a relay. In the proposed scheme, the relay applies algebraic systematic Network Coding (a systematic linear block code) to the detected source codewords and forwards only the additional redundancy to the destination where an overall product codeword is observed. Firstly, we focus on the time-division multiple-access (TDMA) cooperation scheme and we suppose a noisy source-relay channel. We analyze the theoretical error probabilities at the relay input and output for different relay strategies. We also establish a theoretical frame error probability bound for the proposed cooperative network in the case of soft decoding at the relay. Then we investigate the relay error mitigation at the destination using the log-likelihood ratio limitation. We evaluate the multi-relay cooperation to deal with the error correlation in the relay-generated redundancy. Different cooperation schemes (single or multi-relay cooperation using different relay strategies) are compared. We show that the source-relay link is the network bottleneck. Then we evaluate the network capacity in the short code-length regime. Lastly, we study the code-division multiple-access (CDMA) based relay cooperation where the relay signal is superposed with the source signals in the same radio-frequency band. To simply the analysis, we suppose an error-free source-relay channel. In order to reduce the mutual interference, we propose a turbo decoding procedure with interference cancelation. We formulate two kinds of CDMA-based cooperation: TDMA-CDMA and FDMA-CDMA. The optimized energy allocation between the sources and relay is evaluated through simulations.
