Abstract
The Psychophysical basis of pattern recognition. This work proposes a pattern classifier that uses the minimum logarithmic distance between the input and the class reference pattern, as an assignation criterion. We show the reasonings based on the Psychophysical laws, which recommend to use this classification algorithm. By Montecarlo simulation, we have proved the superiority of this classifier over the gaussian classifier in many tasks. Finally, a two class pattern classification experiment was designed so that the assignation into a class or other depended on the use of our classifier or the gaussian one. The thirty three experimental subjects performed a classification like the classification obtained with the minimum logarithmic distance algorithm.
