Ratiometric Chemical Blend Processing with a Neuromorphic Model of the Insect Macroglomerular Complex

We present a dynamical spiking neuromorphic model constrained by the known biology of the insect antennal lobe (AL) macroglomerular complex (MGC) implemented in a field programmable gate array (FPGA). When driven by polymer coated quartz-crystal microbalance (QCM) chemosensors at its input, the dynamical trajectories of the model's projection neuron (PN) output population activity encode the concentration ratios of binary odour mixtures. We demonstrate that it is possible to recover blend ratio information from the early transient phase of QCM responses that would otherwise be difficult to separate directly from chemosensor data using classical approaches. Our results demonstrate the potential of insect-based neuromorphic signal processing methods for the rapid and efficient classification of ratiometrically encoded chemical blends.