Motion Processing based on spatio-temporal receptive fields with biphasic temporal response property T Höppner1, F Hamker2 |
|---|
1Künstliche Intelligenz, Technische Universität Berlin, Germany |
| Decoding and understanding motion starts already in the retinal ganglion cells and in the lateral geniculate nucleus (LGN) as they provide a particular temporal characteristic for the motion selective cells in the primary visual cortex (V1). For natural images the visual information consists of a stream of time-varying brightness values and motion as well as its components, velocity and direction, have to be computed from this stream. Here we introduce a neuro-computational model of LGN that is based on space-time dependent receptive fields with biphasic temporal response properties. The spatial structure of the receptive fields are classic center surround ones as found in LGN cells. This spatial structure is modulated by a biphasic temporal function which has been described in visual areas, e.g. LGN, primary visual cortex (V1). These biphasic neural response properties lead to a complete change of the spatial structure from on-center to off-center characteristic and vice versa. However, different from previous models the temporal characteristic has not been pre-defined by a fixed filter function, but it is the result of the computation induced by the temporal changes. We used this model to fit the spatio-temporal receptive fields of LGN cells with different stimulation protocols. |
| Up Home |
|---|