Method and apparatus for implementation of neural networks for face recognition

A method for real-time object recognition comprising the steps of:

• capturing an input image of an object to be recognized;
• storing a plurality of reference images, each of said reference images being indicative of a cluster of images of an object on record;
• and
• nonlinearly jointly correlating said input image with each corresponding said reference image at a first layer of a two-layer neural network to generate a corresponding correlation signal, and
• combining said correlation signals at a second layer having nonlinearity of said two-layer neural network to provide a combined correlation signal, wherein said combined correlation signal above at least one threshold is indicative that the object to be recognized is the object on record.

A real-time object recognition system comprising:

• image detector means for capturing an input image of an object to be recognized;
• a plurality of reference images, each of said reference images being indicative of a cluster of images of an object on record, said reference images stored on an optical film deposited on a card;
• and
• a two-layer neural network comprising,
• (1) a first layer comprising,
• (a) a light source for illuminating each of said reference images to project corresponding reference images,
• (b) a spatial light modulator receptive to said input image, said spatial light modulator in response to illumination by said light source projecting said input image, and
• (c) detector means for detecting combined corresponding said reference images and said input image to provide a corresponding nonlinear joint correlation signal, each of said joint correlation signals being indicative of a joint correlation between said input image and each corresponding said reference image, and
• (2) a second layer having nonlinearity for combining said correlation signals to provide a combined correlation signal.

A method for real-time object recognition comprising the steps of:

• capturing an input image of an object to be recognized;
• storing a plurality of reference images each of said reference images being indicative of a cluster of images of an object on record, storing said reference images on an optical film deposited on a card;
• at a first layer of a two-layer neural network,
• (a) illuminating each of said reference images to project corresponding reference images,
• (b) illuminating a spatial light modulator receptive to said input image to project said input image,
• (c) detecting combined corresponding said reference images and said input image to provide a corresponding nonlinear joint correlation signal, each of said joint correlation signals being indicative of a joint correlation between said input image and each corresponding said reference image;
• and
• combining at a second layer having nonlinearity of said two-layer neural network said correlation signals to provide a combined correlation signal, wherein said combined correlation signal above at least one threshold is indicative that the object to be recognized is the object on record.