Multi-layer opto-electronic neural network

• US 5,497,253 A
• Filed: 07/30/1993
• Issued: 03/05/1996
• Est. Priority Date: 07/18/1988
• Status: Expired due to Term

• Current Assignee: Northrop Grumman Systems Corporation (Northrop Grumman Corporation)
• Original Assignee: Northrop Grumman Corporation
• Inventors: Reis, James J., Stoll, Harold M.
• Primary Examiner(s): Dzierzynski, Paul M.
• Assistant Examiner(s): Juba, Jr., John
• Application Number:
  US08/099,882
• Time in Patent Office: 949 Days
• Field of Search: 359/4, 359/7, 359/11, 359/29, 359/559, 359/560, 359/561, 382/14, 382/15, 382/155, 382/156, 382/157, 382/159, 382/37
• US Class Current: 359/29
• CPC Class Codes: G03H 1/0005 Adaptation of holography to specific applications holographic optical element G02B5/32; holographic scanner G02B26/106; recognition using holographic mask G06V10/88; holographic memories G11B7/0065, G11C13/042
• CPC Class Codes: G03H 1/0248 Volume holograms
• CPC Class Codes: G03H 1/16 using Fourier transform G03H1/10, G03H1/12, G03H1/14 take precedence; analogue computers G06G, e.g. G06G7/19
• CPC Class Codes: G03H 1/265 Angle multiplexing; Multichannel holograms G03H1/268 takes precedence
• CPC Class Codes: G03H 2001/0066 for wavefront matching wherein the hologram is arranged to convert a predetermined wavefront into a comprehensive wave, e.g. associative memory
• CPC Class Codes: G03H 2260/54 Photorefractive reactivity wherein light induces photo-generation, redistribution and trapping of charges then a modification of refractive index, e.g. photorefractive polymer
• CPC Class Codes: G06E 3/001 Analogue devices in which mathematical operations are carried out with the aid of optical or electro-optical elements optical elements per se G02B; devices consisting of a plurality of solid state components, including light sensitive semiconductor components, formed in or on a common substrate H01L27/14; electro-, magneto- or acousto-optics, non-linear optics G02F1/00; graph reading G06K11/00
• CPC Class Codes: G06N 3/067 using optical means
• CPC Class Codes: G06V 10/895 the filter being related to phase processing, e.g. phase-only filters

Abstract

A pattern recognition apparatus and a method for operating same. The apparatus includes a volume holographic medium (4) having a plurality of Fourier-space volume holograms representing pattern templates stored within. The apparatus further includes a spatial light modulator (1) and a phase encoder (2). The phase encoder has an output optically coupled to the medium by a first Fourier transform lens (3). The spatial light modulator spatially modulates a spatially uniform laser beam (7) in accordance with an unknown pattern. The two-dimensional phase encoder causes the spatially modulated laser beam to be spatially distributed prior to application to the medium. The apparatus also includes a detector (6, 11) having an input optically coupled by a second Fourier transform lens (5) means to an angular spectrum of plane waves generated by the medium in response to the output of the spatial modulator, phase encoder, and first Fourier lens. The detector detects plane waves that correspond to vector inner products generated within medium (4) in response to the unknown pattern. The apparatus further contains a means (12) for nonlinearly processing the output of detector (6, 11) and a means (13) by which the output of nonlinear processing means (12) may be temporarily stored.

Claims

1. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus, comprising:

a volume holographic medium having a plurality of Fourier-space volume holograms representing neural network weight vectors stored within;

means, having an output optically coupled to said medium by a first Fourier transform lens means, for spatially modulating a spatially uniform laser beam in accordance with an unknown pattern;

means, having an input optically coupled by a second Fourier transform lens means to an angular spectrum of plane waves generated by said medium in response to the output of said spatial modulating means, for detecting plane waves that correspond to vector inner products generated within said medium in response to the unknown pattern;

means by which the output of said detecting means are nonlinearly processed in a serial manner;

means by which said nonlinearly processed output is temporarily stored; and

means by which the output of said temporary storage means may be selectively read out and selectively fed back into said spatial modulating means for further processing by the MLOENN as input signals to a next layer of the MLOENN.

2. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said spatial modulating means is geometrically partitioned to permit the independent storage of individual MLOENN layer weight vectors, ν

.sup.(k,i (i=1,2, . . . , N-1;

k=1,2, . . . , K), and the independent excitation of individual MLOENN layers by patterns σ

.sup.(i) (i=1,2, . . . , N-1) such that each ν

.sup.(k,i) and σ

.sup.(i) occupy a same pixel of said spatial modulating means.

3. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said spatial modulating means further includes means for phase encoding the spatially modulated spatially uniform laser beam in accordance with a random, two-dimensional phase encoding function.

4. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said detecting means includes a linear array of photoresponsive elements disposed within and along a correlation plane of said second Fourier transform lens.

5. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said nonlinear processing means has a sigmoidal transfer function with adjustable threshold and saturation points.

6. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 5 wherein said nonlinear processing means is either an electronic look-up table or an electronic saturable amplifier.

7. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said temporary storage means is a digital electronic memory.

8. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said volume holographic medium provides spatial filtering, within the correlation plane, of an electric field distribution, the spatial filtering being provided in accordance with the Bragg selectivity of said volume holographic medium.

9. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said volume holographic medium is comprised of a material selected from the group consisting of iron-doped lithium niobate (LiNbO₃ :

Fe), strontium barium niobate (SrBaNbO₃), photorefractive photopolymers, and photochemical photopolymers.

10. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said spatial light modulating means is comprised of means selected from the group consisting of ferroelectric liquid crystal devices, twisted nematic liquid crystal devices, deformable mirror devices, and magneto-optic devices.

11. A multi-layer opto-electronic neural network (MLOENN) pattern classification apparatus as set forth in claim 1 wherein said detecting means includes a two-dimensional array of photoresponsive elements.

12. A method for storing (K) Fourier-space holograms within a volume holographic medium, comprising the steps of, for K Fourier-space holograms ν

.sup.(k,i) (k=1,2, . . . , N);

loading a weight vector ν

.sup.(k,n) into a spatial light modulator means using a predetermined and fixed lexicographic ordering scheme;

spatially modulating a first plane-wave laser beam in accordance with the loaded weight vector;

multiplying a light pattern transmitted by the spatial light modulator means in accordance with a random, two-dimensional phase encoding function;

applying the phase-encoded light pattern, with a first Fourier transform lens means, to approximately a midpoint of the volume holographic medium so as to generate, at the midpoint, a Fourier transform of the phase-encoded light pattern while simultaneously applying a second plane-wave laser beam, which is temporally coherent with the first plane-wave laser beam, so as to illuminate the medium at an angle ψ

_k to a z₂ -axis of the medium and within a x_2-z2 plane;

forming a weight vector hologram ν

.sup.(k) within the medium;

changing the angle ψ

_k to an angle ψ

.sub.(k+1) ; and

repeating the preceding loading, spatially modulating, multiplying, applying, forming, and changing steps for each k=1,2, . . . , K until all ν

.sup.(k) have been stored in the medium.

13. A method as set forth in claim 12 and including a step of fixing the weight vector holograms within the medium, the medium being formed of LiNbO₃ :

Fe, the holograms being fixed by heating the medium to a temperature of approximately 130°

C. for a period of approximately 25 minutes.

14. A method for generating vector inner products (ƒ

_k.sup.(i)) for accomplishing pattern recognition, comprising the steps of;

loading an unknown pattern vector σ

.sup.(i) (i=1,2, . . . , N) into a spatial light modulator means using a same pre-determined and fixed lexicographic ordering scheme that was initially employed to store a weight vector ν

.sup.(k,i) (k=1,2, . . . , K;

i=1,2, . . . , N) template into a volume holographic medium;

spatially modulating a first plane-wave laser beam in accordance with the unknown pattern vector;

employing a phase encoder means to multiply the spatially modulated first plane-wave laser beam by a random, two-dimensional phase encoding function;

transmitting a phase encoded light pattern representative of the multiplied spatially modulated first plane-wave laser beam from the phase encoder means to a first Fourier transform lens means;

generating within the volume holographic medium, with the first Fourier transform lens means, a Fourier transform of the phase encoded light pattern;

generating, within the volume holographic medium, an angular spectrum of plane waves having amplitudes proportional to vector inner products σ

.sup.(i) •

ν

.sup.(k,i) (k=1,2, . . . , K;

i=1,2, . . . , N) and propagation angles of ψ

_k (k=1,2, . . . , K) which correspond to reference angles ψ

_k of a plane-wave coherent optical signal;

focusing onto a detector means, with a second Fourier transform lens means, plane waves of the angular spectrum of plane waves generated within the volume holographic medium;

reading out, from the detector means, signals representing vector inner products corresponding to the plane waves focused upon the detector means;

serially and nonlinearly processing the output of the detector means;

temporarily storing the nonlinearly processed output of the detector means;

feeding the temporarily stored output of the detector means back into the spatial modulating means for further neural network processing as input signals of a next pattern vector σ

.sup.(i+1) and reading out the temporarily stored output of the detector means as pattern recognition results.