Browse Prior Art Database

Simplified Optical Correlator for Pattern Recognition

IP.com Disclosure Number: IPCOM000079339D
Original Publication Date: 1973-Jun-01
Included in the Prior Art Database: 2005-Feb-26
Document File: 3 page(s) / 42K

Publishing Venue

IBM

Related People

Casler, DH: AUTHOR

Abstract

Many patter-recognition machines derive a signature or vector of features Fj from an unknown input pattern. This vector is then correlated against sets of feature weights Wjk, representing each of the pattern classes k to be recognized. That class having the highest correlation output (Image Omitted) is taken as the identity of the input pattern.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 3

Simplified Optical Correlator for Pattern Recognition

Many patter-recognition machines derive a signature or vector of features Fj from an unknown input pattern. This vector is then correlated against sets of feature weights Wjk, representing each of the pattern classes k to be recognized. That class having the highest correlation output

(Image Omitted)

is taken as the identity of the input pattern.

Fig. 1 shows a simplified optical correlator 10 for performing the summation of Eqn. (1). The output of each of a number of conventional feature detectors, not shown, produces a light intensity in a corresponding one of the light sources
11. If the features are binary-valued, the sources will be either fully on or fully off, but analog feature values may also be used to indicate the degree to which a feature is present. Fiber-optic bundles 12 receive light from respective sources
11. The other ends 13 are shaped into thin horizontal lines facing a weighting film 14. Film 14 has a number of areas 15 which are selectively darkened, as a function of the values of the feature weights Wjk. Areas 15 in any row represent the same feature of different pattern classes, while those in a single column represent different features of the same class.

Ends 16 of the fiber-optic bundles 17 form thin vertical lines, one for each column of film 14. The other end of each bundle 17 is coupled to a photodetector
18. The outputs of detectors 18 thus carry electrical signals, each indicating the degree to which the observed features correlate with the weights for a particular pattern class. The largest output signal is selected by a circuit, such as that shown in U. S. Patent 3,714,465, to indicate which pattern class had the greatest correlation.

In practice, dif...