Signal Processing with Spiking Neuron Model

The article deals with signal processing models that use the representation of signals by spike trains. The “input-output” model of an integrate and fire spiking neuron is considered. The possibilities of using this model for spectral analysis and adaptive filtering are investigated. To solve the problem of adaptive processing of spike sequences, a generalized multi-input vector-matrix model of a spiking neuron is introduced. Based on this model, we derive supervised learning rules to adapt the dynamics of the spiking neuron. Examples of adaptive signal recovery from spike trains, additive noise suppression, and adaptive synthesis of band-pass filters are demonstrated.