An artificial neural network for neural spike classification

In insects, the summed responses of neural activity can be obtained by recording from the exterior of a taste organ (sensillum) of an intact animal. These multiunit recordings are commonly used to understand sensory and behavioral physiology. It is possible to distinguish between the neural spikes produced by these chemosensory neurons using such features as amplitude and shape. We have developed an artificial neural network (ANN) spike classifier which is capable of distinguishing among neural responses of each insect taste organ. The ANN is "trained" on prototypical spikes produced by each of the constituent neurons. It performs very well when compared with conventional optimal methods of template matching and principal components.