A neural network incorporating direct optical imaging
A compact neural network architecture (10) is trainable to sense and classify an optical image (Figs 3, 6) directly projected onto it. The system is based upon the combination of a two-dimensional array (12) of amorphous silicon photoconductors (18) and a liquid-crystal spatial light modulator (20;...
Saved in:
Main Author | |
---|---|
Format | Patent |
Language | English French German |
Published |
28.05.1997
|
Edition | 6 |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | A compact neural network architecture (10) is trainable to sense and classify an optical image (Figs 3, 6) directly projected onto it. The system is based upon the combination of a two-dimensional array (12) of amorphous silicon photoconductors (18) and a liquid-crystal spatial light modulator (20; Fig. 2). Appropriate filtering of the incident optical image upon capture is incorporated into the network training rules, through a modification of the standard backpropagation training algorithm. Training of the network (10) on two image classification problems is described: the recognition of handprinted digits (Fig. 3), and facial recognition (Fig. 6). The network (10), once trained is capable of standalone operation, sensing an incident image and outputting a final classification signal in real time. |
---|---|
Bibliography: | Application Number: EP19950309053 |