Optical data processing techniques have the inherent advantage of high data throughout, low weight and low power requirements. These features are particularly desirable for onboard spacecraft in-situ real-time data analysis and data compression applications. The proposed multi-layer optical holographic neural net pattern recognition technique will utilize the nonlinear photorefractive devices for real-time adaptive learning to classify input data content and recognize unexpected features. Information can be stored either in analog or digital form in a nonlinear photorefractive device. The recording can be accomplished in time scales ranging from milliseconds to microseconds. When a system consisting of these devices is organized in a multi-layer structure, a feed forward neural net with bifurcating data classification capability is formed. The interdisciplinary research will involve the collaboration with top digital computer architecture experts at the University of Southern California.

A new bifurcative optical pattern recognizer (BOPAR) for image and/or data classification using the noise fan-out phenomenon in nonlinear photorefractive crystals is described. The inhomogeneities in photorefractive crystals naturally exits consequential to the doping procedure in making the crystals photorefractive. Scattering occurs when light impinges on the inhomogeneities. The scattered light is in general random in intensity distribution and angular orientation. Therefore such scattered light is considered as annoying noise that tends to reduce the strength of the signal. In this paper, a new and exciting discovery is reported. Contrary to the conventional conviction, we found that certain portions of the fan-out noise can be made useful for achieving bifurcative optical holographic pattern recognition and data classification. The BOPAR has a neuromorphic or brainlike nature.

Trinary associative memory combines merits and overcomes major deficiencies of unipolar and bipolar logics by combining them in three-valued logic that reverts to unipolar or bipolar binary selectively, as needed to perform specific tasks. Advantage of associative memory: one obtains access to all parts of it simultaneously on basis of content, rather than address, of data. Consequently, used to exploit fully parallelism and speed of optical computing.

A liquid-crystal TV spatial light modulator (LCTV SLM) input device and LCTV nonlinear thresholding element are presently used to accomplish an all-optical implementation of an inner-product neural associative memory. This architecture represents an alternative to the vector-matrix multiplication method of the Hopfield model, which is most often employed by neural network associative memory models. LCTV SLM experimental results are presented and discussed.

The convergence mechanism of vectors in Hopfield's neural network in relation to recognition of partially known patterns is studied in terms of both inner products and Hamming distance. It has been shown that Hamming distance should not always be used in determining the convergence of vectors. Instead, inner product weighting coefficients play a more dominant role in certain data representations for determining the convergence mechanism. A trinary neuron representation for associative memory is found to be more effective for associative recall. Applications of the trinary associative memory to reconstruct machine part images that are partially missing are demonstrated by means of computer simulation as examples of the usefulness of this approach.

To increase the storage capacity of the Hopfield-type neural network, the spurious states need to be reduced or eliminated. Recently, a new type of attractor called a terminal attractor, which represents singular solutions of a neural dynamic system, has been introduced1 for the elimination of spurious states in associative memory. These terminal attractors are characterized by having finite relaxation times, no spurious states, and infinite stability. They provide a means of real time high density associative memory applications and potential solutions of learning and global optimization problems. The original terminal attractor model assumed continuous variable representation of neural states in the neural dynamic equations. Also, sigmoidal thresholding functions are assumed. These assumptions present difficulties for optical implementations. To implement the terminal attractor model, we have made modifications to incorporate this model to the associative memory system.

Experimental system demonstrates optical processing of multiple channels of binary signals. One input channel contains signal that varies with time and applied to one-dimensional acousto-optical cell. Other input channel contains two-dimensional pattern that is stationary or can vary with time and applied to magneto-optical spatial light modulator. Output is time-varying correlation or convolution of first input with one of rows in second input.

The papers presented in this volume provide an overview of current research in both optical and digital pattern recognition, with a theme of identifying overlapping research problems and methodologies. Topics discussed include image analysis and low-level vision, optical system design, object analysis and recognition, real-time hybrid architectures and algorithms, high-level image understanding, and optical matched filter design. Papers are presented on synthetic estimation filters for a control system; white-light correlator character recognition; optical AI architectures for intelligent sensors; interpreting aerial photographs by segmentation and search; and optical information processing using a new photopolymer.