METHOD FOR QUANTIZATION USING COMPETITIVE LEARNING METHOD FOR MODULATION OF PULSE CODE

摘要

PURPOSE: A method for quantization using a competitive learning method for modulation of pulse code is provided to perform easily a quantization process by using a competitive learning method of a neural circuit network. CONSTITUTION: A method for quantization using a competitive learning method for modulation of pulse code comprises the following steps. A neural circuit network is initialized and a distance between an input value and a weight connection value is obtained. A real output value is calculated by using a reciprocal number of the obtained distance. An output node having the maximum value of the calculated output values is searched. A connection weight value adjacent to the weight connection value and the output node having the maximum output value and the output node having the maximum output value is updated through competition of each output values. Quantization is finished according to the number of the update.