Representation of Spoken Words in a Self-Organizing Neural Net

This paper will deal with an algorithm for a twodimensional representation of the acoustic signal of spoken words. Having such an algorithm one can use the result for different applications like speech recognition and machine control by voice. If such a map from the domain of continuous speech signals to a two- (or three-) dimensional Euclidean space can be argued to be a reasonable model of the human speech processing system, then one can even investigate with this model psycho linguistic phenomena. The algorithm is in fact a shortcut for a self-organizing artificial neural network as developed by Kohone