Cooperative communication has recently been considered as a key technology for modern wireless standards and next generation wireless networks to improve the quality-of-service (QoS) and extend transmission coverage in a cost-e ective manner. In this thesis, we aim to address the challenges of wireless cooperative communications, and design the e cient OFDM-based relay systems. Starting with the channel accumulation problem in the amplify-and-forward (AF) relay system, we proposed two adaptive guard interval (GI) schemes for single/multiple-relay networks respectively to cover the accumulated delay spread and enhance the transmission e - ciency. Distinct from the traditional adaptive GI, the dynamical GI length can be detected by the destination individually. Numerical results show that the proposed scheme can further save the control signaling overhead without any symbol error rate (SER) performance loss. For multiple-relay systems, a novel relay selection criterion is proposed to achieve the trade-off between the transmission reliability and overhead by considering both the channel gain and the accumulated delay spread. From computer simulations, the proposed relay selection scheme significantly improves the e cient throughput over the multipath channel with variable channel length, Moreover, cooperative systems require an accuracy resource allocation to achieve the high capacity in the two-way decode-and-forward (DF) relay system with time-varying channels and the bidirectional asymmetric tra c. We propose two allocation algorithms, where the total capacity is maximized under a capacity ratio constraint which depends on the tra c-load di erence between the uplink and downlink. The balanced capacity performance shows that the proposed schemes can assure the fairs data rate of the two terminals and improve the overall QoS of the relay network. A low-complexity suboptimal allocation algorithm is proposed for the frequency-division model which separates subcarrier and time/power allocation. Verifying by simulations, the suboptimal scheme can achieve the similar performance with the optimal one with reduced complexity. In order to further enhance the transmission reliability and maintain low processing delay, we propose a novel equalize-and-forward (EF) relay scheme, which can equalizes the channel between source and relay and eliminate the channel accumulation e ect. The relay processing time is reduced by performing the channel estimation and equalization in parallel. In the EF relay, free-delay equalization is realized by presetting the equalizer with the current channel response that is predicted in parallel. Numerical results show that the EF relay can achieve comparable SER performance as the DF relay with much less latency and exhibit low outage probability at the same data rate as compared to traditional AF and DF schemes.