Figure 3.

Illustration of signal pre-processing to convert DVS events into equivalent analog input for ESNs. In order to reduce the number of input channels to the system and reduce computational load, the input from retina pixels is first spatially subsampled into cells C(x_c, y_c). Each set of ESNs then receives input from a particular receptive field RF(x, y, t₁, t₂). To compute the equivalent analog input, an exponential kernel finally is applied to each event contained in the receptive field.