Modelling Pattern Recognition in Cricket Phonotaxis

Abstract

A spiking neuron implementation of pattern recognition of the calling songs in Gryllus bimaculatus is proposed. A simplified model of the auditory interneuron AN1 has been fitted to extracellular physiological data. The model captures the aspects of AN1’s rate-response to acoustic stimulation which are believed to be sufficient for pattern recognition. Stimulation patterns can be induced into the model via current injecton of the signals envelope-shapes. The model was used as the input stage to the pattern recognition mechanisms. A biologically plausible filter mechanism for pulse-pause patterns is proposed which is based on short term synaptic plasticity. Three simple filter mechanism are described, based on either isolated synaptic depression or synaptic facilitation. These filters are able to reproduce physiological findings from the cricket’s auditory brain neurons. Further, it is argued that more complex filters can be produced by using combinations of depression and facilitation, and that a complete model of the cricket’s pattern recognition apparatus may be implemented in this way. This however is left as a subject of further studies.