Dynamic coupling weights in a neural network system

A number of applications of Neural Network Systems (NNS) for speech signal processing have been developed, Kohonen (1988), Wu, Warwick and Koska (1990), Wu and Warwick (1990). The concept of feature maps makes the display of phoneme recognition result directly from the output of a NN System in the form of a two-dimensional map. Chinese phoneme feature maps have also been tested in a similarly structured NN System. The authors introduce new improvements in that the dynamic coupling weight concept is provided by means of a switch structure. This makes the weights changeable even during the recognition process of the NN system. The required function can be achieved to enhance the feedback between units either in a positive or negative fashion, furthermore, a multi-switch structure can also be designed. This enlarges the adjustable range of weights, and above all, the improvement provides the opportunity to use knowledge from an artificial intelligent system to control the logic of the switches.<
>