Technical: Audio Analogy
It has been seen that almost all of the information derived by the human visual sense is obtained, not by Nyquist sampling of the input optical image, but by low density SD sampling of the spatial frequency "power" within the input image.
In the human aural sense all of the information is obtained, analogously, not by Nyquist sampling but by low-density sampling of the temporal frequency power within the input audio.
Automatic speech recognition systems operate with the same type of information that is sampled by the aural sense, not on Nyquist sampling of the audio input.
Correspondingly automatic pattern recognition should be based upon the information that is sensed by human vision, not on Nyquist sampling of the input image.