The current study is part of a larger project aiming at offering intuitive mappings for the control of synthesis models by semantic descriptions of sounds, i.e. simple verbal labels related to various feelings, emotions, gestures or motions. Hence, this work is directly related to the general problem of semiotics of sounds. We here put a special interest in sounds evoking different perceived motions. To focus on intrinsic invariants of sounds, we have adopted the "acousmatic" listening approach by constituting a set of sounds composed of recorded sounds for which the sound producing sources are as unrecognizable as possible. We also included synthesized sounds to examine specific assumptions related to the physics of moving sound sources. We then studied the perceptual categorization of these sounds using categorization tasks. In this paper, the experimental design of the listening tests is described and the results obtained from behavioral data are discussed. We finally present some perspectives directly linked to synthesis applications.