Neural network device and image recognition method employing photoconductive liquid crystal device with patterned electrode

A light modulation device comprising:

• a first transparent electrode layer, said first transparent electrode having a uniform voltage applied thereto;
• a photoconductive layer formed on said first transparent electrode layer;
• said photoconductive layer exhibiting an increase in electrical conductivity when exposed to light;
• a light modulation layer formed on said photoconductive layer for modulating a read light in accordance with an input image incident on said photoconductive layer;
• a second transparent electrode layer formed on said light modulation layer;
• and
• a conductive electrode formed between said photoconductive layer and said light modulation layer;
• wherein said conductive electrode is divided into an n.times.m matrix of regions each corresponding to pixels of a digitized light input pattern, said conductive electrode forming a single layer,
• a plurality of unit electrodes electrically isolated with each other are arranged in each region except for regions on the outer border, and each of said plurality of unit electrodes in each region is electrically connected to one of said unit electrodes in at least one of regions surrounding each said region, wherein a read light incident on said light modulation layer and said conductive electrode is modulated at regions having unit electrodes electrically connected with each other when a summation of outputs from said unit electrodes electrically connected exceeds a predetermined threshold value.