In wireless multi-hop networks, hello protocols for neighbor discovery are a basic service offered by the networking stack. However, their study usually rely on rather simplistic models which do not take into account problems resulting from low level layers, such as the physical layer. One of the peculiarities of radio communications is the presence of interferences which decrease the capacity of the medium.In this paper, we consider a random hello protocol inspired by aloha and we study the impact of the interferences on the neighbor discovery process. As expected, we prove that, in average and in the presence of interferences, a node discovers only a subset of its neighbors. We propose then an analytical model to compute the average number of nodes that a given node may expect to discover in its neighborhood.Finally, we present a hello protocol with sleep periods. We show how to optimize this protocol using our hybrid model. A real scenario stemming from the CAPNET project is then analyzed and studied.