Decoding high-frequency components of the input signal from slowly spiking simulated neurons

Abstract

It is well known that neurons in the sensorimotor cortex of primates tend to synchronously activate and generate a pulsating motor command, and such rhythmic cortical drive induces voluntary contraction of muscles. Using frequency analysis, numerous experimental data have revealed that even a single neuron in the motor system transmits a rhythmic signal with a frequency much higher than its own spike rate, although this seems odd when we consider the sampling theorem. To prove such physiological findings, we investigated the transmission mechanism of a neural signal using both mathematical and simulation approaches. For the mathematical consideration, we sampled an analytical signal as both an evenly spaced and unevenly spaced time series in order to simulate regularly and irregularly spiking neurons. The results of the discrete Fourier transform suggest that only data sampled at irregular intervals contained the exact spectrum of the input signal under the condition where the sampling rate was even lower than the input frequency. By observing the signal structure, we concluded that harmonics-cancellation was achieved due to irregular sampling intervals. We also simulated both regularly and irregularly spiking neurons using Hodgkin. Huxley's neural membrane model. When we applied sinusoidal current inflow of 135.0 Hz to the neuron, the neuron spiked with constant intervals (this case is comparable to the regularly sampled data). In this case, the spectrum did not reflect the input signal. However, in the case of sinusoidal input with frequency of 132.0 Hz, the spectrum showed the input frequency content. The neural spike rate (that is, the output frequency), was much lower than that of the input frequency. From these results, we concluded that fluctuation of spiking in neural output might play an important role for transmitting information in the input signal.