A neural network, which is capable of recognizing isolated spoken numbers speaker independently will be described. The recognition system is hybrid of self-organizing map (SOM)and multilayer perceptron (MLP) neural networks. The SOM maps the feature vectors of a word into a constant two dimensional binary matrix. Which is classified by an MLP. The mapping of a SOM is visualized by cumulating these binary patterns. The decision borders of the SOM were fine-tuned with LVQl algorithm, with which the hybrid achieved over 99
展开▼