Artificial neural networks (ANN) have been inspired by the human brain information processing techniques has recently found its place in many engineering applications. In particular, classification, modeling and prediction applications of ANN have attracted quite attention because ANN provides high performance. The wireless cooperative communication, a communication system in which users are communicating through a relay destination. In this paper, error performance of ANN based cooperative communication technique is studied. This technique, which employs training sequences, is a prominent approach for decreasing the system complexity and cost by eliminating the need for channel estimation. The numerical results for Rayleigh fast fading channels over BPSK modulation have shown that proposed approach provides full diversity gain.