Analog CMOS current-mode implementation of the feedforward neural network with on-chip learning and storage

Based on the outstar structure and the ratio memory, a feedforward Hamming neural network with on-chip learning and storage is designed in CMOS current-mode circuits. The implemented feedforward net can be used as a pattern classifier. The chip of the feedforward Hamming net has been fabricated by 0.8 /spl mu/m double-poly double-metal n-well CMOS process. The experimental results show that the ratio memory has the contrast enhancement characteristic. Also, the classifier can recognize the distorted pattern which is darken, brighten, level-shifted, or noisy of the exemplar pattern with gray level recognition capability.