The aim of this paper is to present a new concept for target motion analysis within a binary sensor network. For the sake of simplicity, we focus on a constant target motion. The binary information represents a very rough information about the perception of the target motion by an elementary sensor, i.e. is the target approaching or going away. Collecting these binary informations, a first step is to determine the information we can extract at the network level about target motion. Then, based on this step, new concepts are introduced for infering the target motion parameters. One is based upon the separation properties and relies on the SVM formalism; while the other one uses the concept of the velocity plane and the PPR (Projection Pursuit Regression) framework. Moreover, theoretical results about the convergence of this method are also presented.