Leaky integrate-and-fire neurons + STDP. A neuron sums incoming spikes but leaks between them — so spikes that arrive together push it over threshold and it fires, while the same spikes spread out leak away and it stays quiet. That's information in the timing — which a rate model throws away. Slide the spread; then train it to classify temporal patterns.
Five distinct spatiotemporal spike patterns, one per class. Train, and STDP shapes the synapses until each pattern lights its own output neuron.