Coincidence Detection Using Spiking Neurons with Application to Face Recognition

Figure 14

Different spike codings applied to output spikes of CD classifier fully trained by Prediction with epoch = 1, taking inputs from AR Scream S1 on the left and AR Scarf S1 on the right. Figures in row (a) are plotted from matching face samples, while in bottom row (b) they are from nonmatching face samples. Scatter plots are comprised of spatiotemporal input spike patterns of reference face (blue asterisk) and test face (red circle) from 7056 neurons afferents while bar plots on the right show the CD neurons latency, rate, rank order, and TTFS coding applied. The input spikes are obtained by direct conversion of pixel values to time domain which is then normalized to be between 0 and 18.

(a)

(b)