Waveform recognition and classification using an unsupervised network

This article describes the modified version of a neural network which learns by experience. It is enhanced with the analog signal processing capability. Basically, it is a kind of unsupervised learning neural networks, with two operating parameters-learning rate and mean square error. By using different values of the parameter, mean square error, this network can provide either an one-to-one mapping (recognition) or a clustering function (classification). Also, it needs less memory storage and takes less steps to provide stable outputs as compared with other unsupervised neural networks. This network can also accept discrete data with continuous values, so now it can recognize patterns encoded with continuous values instead of binary values. Hence, it needs much less input neurons as compared to the binary ones when dealing with patterns of the same complexity.