All-optical Pattern Identification System Of Optical Neural Networks Based On Fractional Fourier Transform

Optical neural network mathematical model first proposed byHopfield in 1980, followed by Psaltis and Farhat reported forthe first time in 1985 based on optical vector - matrix multiplication,Hopfield neural networks, optical neural network was completedonthe basis of Ohta et al.miniaturization of integrated chip, from theopticalmethod, the prelude to the neuralnetwork wasopened. After a lot of literature search found that the Fourier transform-basedoptical neural network feature recognition there are many,butbased on the application of basic theory of the fractional Fouriertransform optical artificial neural network feature recognition isstill no systematic and comprehensiveintroducton.This paper is based on the Hopfield optical neural networkmodel and based on fractional Fourier holiographic memory storage,and based on the scoresrelated to the fuction of pattern recognition,given an experimental model of the optical neural network patternrecognition, given an experimental model of the optical neuralnetwork pattern recognition, in order to construct a real-timeidentify ability, high precision , all-optical pattern recognitionsystem, using optical methods to achieve the memory of associative recognition function of the human brain. Obtained in theresearch process through analysis of the principles of thetheoretical results of pattern recognition, in order to verify theconclusions of there ability and feasibility of the experiment usingthe computer simulated results show that: 1)the fractional powerspectrum is a anamorphic fractional correlation，which ismodulated by a power phase factor，the spatial displacement isrestrainted by sinφ₂，and the phase delay is restrainted by the cosφ2.2)The dependent relationship between the feature recognition andthe fractional Fourier transform’s p2order is that when p-2∈[0，1]，thespatial displacement monotonous increase；when p₂∈[0，2]，thephase delay monotonous decrease. And when p₂=1/2，the effect ofspatial displacement and phase delay are equal. Although the resultof anamorphic fractional correlation have no influence of theintensity of the output of the fractional correlation，the relativefact between A and b，f，p₁，p₂, dispersion shows fractionalcorrelation’s variability and space-frequency joint appearance’sinformation extraction ability. It will supply more basic applicationtheory for fractional correlation and fractional Fourier transform.