Neural circuits closer to the periphery tend to be organised in a topological way, i.e. stimuli which are spatially close tend to be mapped onto neighbouring processing neurons. The goal of this study is to show how motion features (optic-flow), which have an inherent spatio-temporal profile, can be self-organised using correlations of precise spike intervals. The proposed framework is applied to the spiking output of an asynchronous dynamic vision sensor (DVS), which mimics the workings of the mammalian retina. Our results show that our framework is able to form a topologic organisation of optic-flow features similar to that observed in the human middle temporal lobe.
Self-organisation of motion features with a temporal asynchronous dynamic vision sensor
Read the paper