Theoretical analysis of a TIME-FREQUENCY-PCNN auditory cortex model.
A particular pulsed neural model of the auditory cortex, the Time-Frequency Pulse Coupled Neural Network (TF-PCNN), has shown to decompose its stimulus from the cochlea into characteristic pulse-coded time-frequency (TF) stop and pass regions. A derived sum of spectrograms representation with these zero- or one-valued TF weightings has already been applied for the denoising of speech in a previous work. This decomposition is now related to the concept of TF projection filters, which allows to reinterpret the model equations. The functionality imposed by the model equations can so be accessed and justified from a TF signal processing perspective, in addition to its biological motivation from experimentally observed neurophysiological behavior and simulations.