JPS63307438A - Optical associative storage device - Google Patents

Abstract

PURPOSE:To shorten a processing time to continuously output the output by converting an input or feedback signal to an optical image and multiplexing/ expanding and forming the image on the photoelectric surface of a space optical modulating tube and performing the data processing by optical operation. CONSTITUTION:The input from a CPU 351 is written on a space optical modulating tube (T) 301 through pattern presenting devices 361 and 362 and expanding/ multiplexing image forming systems 311 and 312. Shutters S1 and S2 are opened and closed in a prescribed order to perform a prescribed operation through Ts 301, 302 and 303 by a monochromatic light, and the result is inputted to the CPU 351 from a reverse multiplexing image forming system 313. The CPU 351 performs the threshold processing and the feedback processing to complete inscription. Remembrance is performed through the system 312 by a prescribed operation in Ts 301, 302 and 303. Thus, the data processing time is shortened and the output is continuously outputted.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to an optical content addressable memory device.

[Conventional technology]

-a. In normal memory used in electronic computers, etc., a method is adopted in which the information stored there is accessed by specifying an address, but in such a storage device, the contents of the address There is a problem in that if the memory is destroyed, the contents of the stored data will become completely unknown. In response, an associative memory device has been developed that finds information that is the same as a reference input applied from the outside or that meets certain conditions, and extracts the remaining related parts.

[Problem that the invention seeks to solve]

Most of these associative memory devices were developed as storage devices for computers, target electrical digital signals, and are constructed using electronic circuit technology using integrated circuits. . Particularly for pattern information, scanning operations are required and large-scale parallel processing is required, resulting in very long processing times, and it is almost impossible to obtain continuous output. Since the circuit configuration involves arranging a large number of arithmetic circuits, it becomes extremely difficult to interconnect them.

The present invention is intended to solve the above-mentioned problems.By performing data processing using optical calculations, the circuit configuration can be made into a front-vehicle configuration, and the processing time can be shortened. The purpose is to provide an associative memory device.

[Means for solving problems]

To this end, the optical content addressable memory device of the present invention includes an optical device that presents an input or feedback signal, and an enlarged imaging optical system that enlarges and images the output from the optical device on the photocathode of the first spatial light modulation tube. , a multiplex imaging optical system that multiplex images the output from the optical device onto the photocathode of the second spatial light modulation tube, and reads out the first spatial light modulation tube and the second spatial light modulation tube in this order. an imaging optical system that images the multiplication output or the output of the second spatial light modulation tube on the photocathode of the third spatial light modulation tube, the second spatial light modulation tube, and the third spatial light modulation tube; An inverse multiplex imaging optical system that inversely multiplexes and images the multiplication outputs executed by sequentially reading out the modulation tubes on the light receiving device, performs dark value processing on the output from the light receiving device, and generates a feedback signal and human power. It is characterized by comprising a CPU that processes signals and the like.

[Effect]

The optical content addressable memory device of the present invention converts input or feedback signals from the CPU into an optical image, and writes enlarged and multiplexed images of this optical image to first and second spatial light and modulation tubes. , the output of the multiplication operation executed by sequentially reading out the first and second spatial light modulation tubes, or the output of the second spatial light modulation tube is written into the third spatial light modulation tube, and the second,
The multiplication outputs executed by sequentially reading out the third spatial light modulation tube are demultiplexed and photoelectrically converted, and this electrical output is subjected to dark value processing by the CPU, thereby combining the optical system and the electrical system. It becomes possible to realize an optical content addressable memory device.

〔Example〕

Examples will be described below based on the drawings.

First, the principle of the optical associative memory device according to the present invention will be explained.

In the present invention, an autocorrelation matrix is used as a method for realizing associative memory, and a memory matrix is formed by the autocorrelation of the contents to be stored.

The arithmetic processing during storage can be expressed as follows.

M=Σχ・r ・・・・・・・・・・・・・・・・
(1) Here, χ is an input vector representing the content to be stored, r is a transposed vector of χ, and M is a storage matrix. I! IT takes the autocorrelation of the content you want to remember and adds it up many times to form it.

At the time of recollection, the whole can be recalled from a part by calculation with this memory matrix. The calculation process during recall can be expressed as the following formula.

V error φ (M・χ) ・・・・・・・・・・・・・・・
(2) Here, V is the output vector, χ is the input data, and φ is the dark value operation. By operating M・χ, χ
Even if the data is incomplete due to missing parts,
If the memory matrix M is created by the calculation process of equation (1), then
As the recalled data V, data close to the original data χ with the missing parts filled in is obtained. Note that the noise portion is cut by detecting data above a predetermined level by operating the threshold (fi) of φ.

Next, as a method for when the recollection using the memory matrix in equation (1) above does not provide sufficient separation, a sequential calculation method for forming a memory matrix with a higher degree of separation will be described.

M @ 61 ”” M @+α(χ−φ(M, 1 ・
χ))〆・・・・・・・・・・・・・・・・・・(3)
Here, α is the learning gain. The n+1st memory matrix M□ is the result of recalling by M7 the nth time.
It is obtained by correcting M7 by multiplying the correlation between the error component at the time of recall indicated by the difference between (Mll ·χ) and χ and K by the learning gain α. Note that the learning gain α is selected to a value such that MR converges.

In this way, by performing the calculation process of equation (3) until M7 converges, a correlation matrix M with an improved degree of separation is obtained, a memory matrix is formed by the autocorrelation of the contents to be stored, and when associating, this memory 1 + By correlating with columns, you can recall the whole from a part.

FIG. 6 is a diagram for explaining the configuration and operation of a spatial light modulation tube, which is a basic component of the optical associative memory device according to the present invention, in which 1 is a human image, 2 is a lens 3 is a photocathode, 4 is is a microchannel plate 5 is a mesh voltage, 6 is a crystal, 61 is a charge storage surface, 7 is a half mirror, 18 is a monochromatic light,
9 is an analyzer, and 10 is an output image.

In the figure, a human image l incident on a photocathode 3 of a spatial light modulation tube via a lens 2 is converted into a photoelectron image. After this photoelectron image is multiplied by the microchannel plate 4, a charge pattern is formed on the charge storage surface 61 of the crystal 6. The electric field across the crystal 6 changes depending on the charge pattern, and the refractive index of the crystal 6 changes due to the Bonkels effect.

Here, when the crystal 6 is irradiated with linearly polarized monochromatic light 8, the polarization state of the reflected light from the charge accumulation surface 61 has changed due to the birefringence of the crystal 6, so if it is passed through the analyzer 9, An output image 10 having a light intensity corresponding to the light intensity of the input image 1 is obtained.

Next, the main functions of such a spatial light modulation tube related to the present invention will be explained.

(b) Memory function The spatial light modulation tube has a memory function that retains the charge on the surface of the electro-optic crystal for a long time. In conclusion, since the crystal 6 has a very high electrical resistance value, the charge distribution on the crystal surface 61 can be maintained for several days or more.

(b) The subtractive function spatial light modulation tube can selectively form a positive or negative charge distribution on the surface of the electro-optic crystal.

FIG. 7 is a graph showing the secondary electron emission characteristics of an electro-optic crystal.

As shown in FIG. 7, if the primary electron energy E incident on the charge storage surface 61 is smaller than the first crossover point E1 or larger than the second crossover point E2, the number of primary electrons in the crystal Since the number of secondary electrons emitted at the surface is larger (δkl), the crystal surface becomes negatively charged. 1
When the energy of the secondary electrons is between El and E2, the number of secondary electrons is greater than the number of primary electrons (δ>1), so the crystal surface is positively charged.

Writing in this band using positive and negative potentials is performed by controlling the voltages Vc and vb shown in FIG. Either write a negative voltage here first, then charge it positively, or
Alternatively, the subtraction function can be provided by two methods: first charging positively and then writing negative charges. The amount of subtraction can be controlled in three ways:

11μ, a method of changing the intensity of the incident light during subtraction, a method of changing the duration of the voltage applied to the microchannel plate 4, and a method of changing the voltage applied to the microchannel plate 4.

(c) Real-time threshold operation function The spatial light modulation tube can perform real-time threshold operation according to the setting conditions of Vc and vb shown in FIG. In an actual FJ, when the mesh voltage Vc is set as low as about 0°IKV and vb is lowered to 0.IKV or higher, the crystal surface becomes a negative potential and electrons no longer reach it. However, when vb is slowly lowered in a ramp-like manner, electrons corresponding to the intensity of the incident light reach the crystal surface, and the potential of the crystal surface decreases. There are parts that have a negative potential and are not written, and parts that do not have a negative potential and are written, and as a result, threshold operation has been performed depending on the intensity of the incident light.

(d) AND operation based on real-time lJ value adjustment operation The afa spatial light variable m tube has a real-time threshold operation function, as described above. Using this, an AND operation can be performed as follows. ! II. As shown in FIG. 8, when two inputs are superimposed, the input light intensity becomes three levels of 0.1.2. If a real-time threshold operation is performed at level 2 of these, the result will be as shown in FIG. 9, which means that an AND operation has been executed.

Next, other components necessary for the present invention will be explained. Note that the input image is assumed to be a 2×2 pattern.

(2) Enlarged imaging system As shown in FIG. 10, a 2×2 input image is enlarged and projected into a 4×4 image. In reality, this is done using a lens or the like.

(2) Multiple imaging system As shown in FIG. 11, a 2×2 input image is repeatedly projected. Actually, it is realized by a diffraction grating or the like.

(2) Inverse multiplex imaging system As shown in FIG. 12, 4×4 input images are projected so as to be superimposed on a 2×2 image. In reality, this is realized using diffraction gratings, gratings, etc.

Next, a coding method suitable for optical calculation using the aforementioned magnification imaging, multiplex imaging, and inverse multiplex imaging is used.
This will be explained as image No. 2.

The memorization and retrieval method that has been separately filed by the present applicant will be explained.

The vector χ and its transposed vector r are expressed as follows.

In this case, memory is expressed as , while recollection is expressed as χ5XaXaJLyaJ because j2aZt Xsχ2 .

However, in order to perform image processing with this method,
It is necessary to convert two-dimensional image data into one-dimensional data, and this calculation must be electrically processed.

In contrast, in the present invention, storage and recall are performed by the following calculations. Let the content to be stored now be χ, and the following is performed.

Here, by the enlarged imaging system, (χ3 χsea mj) is obtained by the multiple imaging system. At this time, if the multiplication of the elements of χwag and χmlt, χsag ・χ−1t, is M, then (・−・・−・
The elements are multiplied together, and the same result as that obtained by the matrix operation described above is obtained (however, the array is different), and χmag ・
It turns out that it is possible to memorize using χ-It.”
Karu.

Also, let , and the multiplication M−vmlt of the elements of M and ywlt is l
', then z'wMHl/mlt, where l is the inverse multiplex imaging of 1, then
It can be seen that the result is the same as the result obtained by the matrix operation described above (however, the arrangement is different), and that a recollection pattern can be obtained by inverse multiplex imaging of M-ymlt.

In this way, it is possible to perform routing using a multiplex imaging system and an enlarged imaging system, and recall using an inverse multiplex imaging system, making it possible to realize associative memory entirely through optical processing. It becomes possible.

Next, the optical associative memory according to the present invention will be specifically explained. Note that the calculation is performed using the following equation, which is similar to the above-mentioned equation (3).

Mll, 1=-M, +α(χsag -φ(Mn・χI
Ilt)wag)ymlt・・・・・・・・・
...(4) Here, α is the learning gain, 718g is the enlarged imaging of χ, and χmlL is the multiple imaging of χ.
” The first memory matrix M□1 is the result of recall by the n-th M, φ(M, ・ymlt,) W
It can be obtained by correcting M by the product of the error component at the time of recall indicated by the difference between ag and 718g and χ+wlt multiplied by the learning gain α. As mentioned above, the learning gain α is selected to a value that makes M9 converge, and if this calculation process is performed until M7 converges, M7 with improved separation is obtained.
is required.

FIG. 1 is a block diagram showing the configuration of an optical content addressable memory device according to the present invention.
05 is a second arithmetic unit, 106 is an inverse multiplex imaging system, and 107 is a dark value memory.

In the figure, as mentioned above, the enlarged imaging system 101 creates 718g of 16×16 from the 4×4 input χ,
The multiplex imaging system 102 similarly receives 16X1 inputs from 4×4 input χ
Create a χ+alt of 6. The first arithmetic unit 103 is 71
Multiply the elements of 8g and ymlt. The first memory 104 is a memory that performs addition of χsag·χ@ll and subtraction of feedback values. Second computing unit 10
5 multiplies the contents of the memory 104 by ymlt to create recall data.

The demultiplexing imaging system 106 converts 16×16 optical data into 4×
-4 optical data. The dark value memory 107 is a memory that performs a threshold (a operation) to detect data above a predetermined level and cut noise. The output V of this threshold memory is fed back to the input side through a feedback system.

Next, the operation during storage will be explained. Note that the flow during storage is shown by thin lines in the figure.

■First, the second arithmetic unit 105 reads M from the memory 104.
7 and ymlt from the multiplex imaging system 102, M, l
- Calculate ymlt, perform threshold value operation in the threshold value memory 107 via the inverse multiplex imaging system 106, and then store it in memory.

■ From the input data χ to be stored, use the enlarged imaging system 101 to
ag is created, and at the same time, ymlt is created using the multiplex imaging system 102.

■Multiply the elements by the first arithmetic unit 103 using the created χsag and ymlt as input data χ-ag ・χw
M stored in the memory 104 is multiplied by α.
, add to.

The results of ■■ are returned to the input by the feedback system, and enlarged and imaged by the enlarged imaging system 101 (M, ・ymlt).
Let's say Monag.

■In the first arithmetic unit 103, (M, l・ymlt) sa
g ・χ−It is calculated and multiplied by α to M in the memory 104,
Subtract from l.

Input data wiring is performed by performing the above operations until M converges.

Next, the time of recall will be explained.

In this case, it is assumed that the storage matrix M has already been calculated and wired. The flow during recall is shown by the thick line in the figure.

■First, from the input data χ, the multiple imaging system 102
Create wit.

■The second arithmetic unit 105 calculates the multiplication with M read out from the memory 104 and M·χ*It, inversely multiplexes the result in the inversely multiplexed imaging system 106, and operates the threshold value in the threshold value memory 107. memorize it.

(2) Read the output from the dark value memory LOT, and if the recall is insufficient, the output is returned to the input via the feed bank system, and the data is recalled again to read out the desired data.

In this way, wiring and recall can be performed.

Next, an embodiment in which the optical content addressable memory device shown in FIG. 1 is constructed using a spatial light modulation tube will be described.

FIG. 2 is a diagram showing an embodiment of an optical associative memory device using a complete optical system using four spatial light modulation tubes.
201 to 204 are spatial light modulation tubes, 211 is an enlarged imaging system,
212 is a multiplex imaging system, 213 is an inverse multiplex imaging system, 221~
228 is a half mirror, 1231 to 238 are reflecting mirrors, 31
．． S2 and S3 are shutters.

In the figure, the computing units 103 and 105 in FIG.
04 consists of a spatial light modulation tube 203 and a dark value memory 107
is composed of a spatial light modulation tube 204.

The operating method will be explained below in order. Note that the calculation is the first
This is performed based on equation (4) as in the figure.

■First, the input U is guided to the magnifying imaging system 201 and 202 via the half mirror 228 and the reflecting mirror 238, respectively, and the U
ag, create Umlt, spatial light modulation tube 201.2
Write each to 02.

■Open shutter S1, half mirror 221, 222,
The spatial light modulation tube 202 → 203 is read out using monochromatic light through the reflecting mirror 231 and the half mirrors 223 and 224, and the M
Calculate M-Umlt and send the result to Half Mirror-2
25. Writing to the spatial light modulation tube 204 through the inverse multiplexing imaging system 213 while performing dark value manipulation by real-time threshold operation.

■ Open shutter S2 and similarly reflect a23 with monochromatic light.
2. By reading out the spatial light modulation tube 201 → 202 through the half mirrors 223 and 224, the Umag-U
mlt is calculated, and half mirror 225, reflecting mirror 233.
234 and 235 to the spatial light modulation tube 203.

■After erasing the spatial light modulation tube 201, open the shutter S3,
Monochromatic light is read out from the spatial light modulation tube 204 through the half mirror 2211 reflecting mirror 236 and the half mirror 226, and enlarged and imaged by the magnifying imaging system 211 via the half mirror 227 and reflecting mirrors 237 and 238 to perform spatial light modulation. tube 201
Write sag to (M, ・UslL).

■Open the shutter S2 and read out the spatial light modulation tube 201→202, (M, -Umlt) sag -U
mlt is calculated and subtracted from the spatial light modulation tube 203. At this time, the result does not become smaller than 0 due to the effect of Mesh.

The wiring is completed by performing the above operations until convergence for each input cover pattern.

Next, the recollection operation will be explained in order.

(2) First, Umlt is written from the input U to the spatial light modulation tube 212 via the half mirror 228 and the multiplex imaging system 212.

(2) Open the shutter S1, read out the spatial light modulation tubes 202-203, calculate M7.Umlt, and write the result to the spatial light modulation tube 204 through the demultiplexing imaging system 213 by real-time M value operation.

■ Read out the spatial light modulation tube 204 and half mirror 227
Get more output. It is also possible to re-recall this output by returning the shutter S3 using a feedback system.

FIG. 3 is a diagram showing an embodiment of a hybrid optical content addressable memory device in which a part of the optical content addressable memory device using the complete optical system of FIG. 2 is replaced with an electronic circuit. In the figure, 30
1 to 303 are spatial light modulation tubes, 311 is an enlarged imaging system, and 31
2 is a multiplex imaging system, 313 is an inverse multiplex imaging system, 321-32
4 is a half mirror; 1331 to 335 are reflecting mirrors; 341 is a light receiving matrix; 351 is a CPU; 361 and 362 are pattern presentation devices; Sl and S2 are shutters.

In this embodiment, the input part and the feedback part are
Replacing it with U facilitates the interface with electronic circuits. Additionally, the number of spatial light modulation tubes used is reduced to three, resulting in a system that can be implemented more easily.

The operating procedure is exactly the same as that shown in FIG.

As described above, in the systems in the two embodiments, multiplication is performed using two spatial light modulation tubes 201 and 202, and 301 and 302, but since it is actually a binary multiplication of (0° ■), , is the same as the logical AND operation. Therefore, by using an AND operation using real-time dark value operation for this part, it is possible to reduce the number of spatial light modulation tubes by one. - Figure 4 shows that by introducing real-time IJ411 operation AND operation, one spatial light modulation tube is frozen and three
This is a diagram showing an example of an optical associative memory device using a complete optical system realized in this book. In Figure 0, 401 to 403 are spatial light modulation tubes, 411 is an enlarged imaging system, 412 is a multiplex imaging system, and 413 421 to 430 are half mirrors, 5431 to 437 are reflecting mirrors, Sl, S2, S3, S
4 is a shutter.

In the figure, the spatial light modulation tube 401 performs an AND operation,
Spatial light modulation tube 402 functions as a memory, and spatial light modulation tube 403 functions as a dark value memory.

First, I will explain about memory.

■ Open shutter Sl and input U, multiplex imaging system 412
multiple images are formed and ULT is written in the spatial light modulation tube 401.

■Read out "Spatial light modulation tube 401-402" and read "M".

- Calculate tJmlt and write it to the spatial light modulation tube 403 through the demultiplexing imaging system 413 by real-time threshold operation while performing threshold operation.

(2) After erasing the spatial light modulation tube 401, open the shutters S1 and S2, create multiple imaging and enlarged imaging, and execute an AND operation of Ue*ag and Umlt by real-time hard clipping.

■Read out the spatial light modulation tube 401 and read out the Umag-Us
lt is added to the spatial light modulation tube 402.

■After erasing the spatial light modulation tube 401, open the shutters S1 and S4, and perform real-time dark value operation to
) executes an AND operation of a number of animals and Us(t).

■Read out the spatial light modulation tube 401, and read out the spatial light modulation tube 402.
Subtract more. At this time, the result is O due to the effect of Mesh.
It doesn't get smaller.

Storing is completed by performing the above operations until convergence for each input pattern.

Next, recall will be explained.

(2) Open the shutter St and write the multiplexed imaging zero wheel 11 of the input U into the spatial light modulation tube 401.

■Spatial light modulation tube 401-402 and readout, M7/Um
lt is calculated, and the result is written into the spatial light modulation tube 403 by real-time IJ(a operation) through the inverse multiplexing imaging system 413.

■Spatial luminosity Read out iFl tube 403 and half mirror 4
Obtain the output from 25. In addition, this output is sent to the shutter SL,
You can also open S3 and return it using the feedback system to recall it again.

FIG. 5 is a diagram showing an embodiment of a hybrid optical content addressable memory device in which a part of the optical content addressable memory device using the complete optical system of FIG. 4 is replaced with an electronic circuit.
1.502 is a spatial light modulation tube, 511 is an enlarged imaging system, 51
2 is a multiplex imaging system, 513 is an inverse multiplex imaging system, 521-52
3 is a half mirror; 1531 to 534 are reflection prints; 541 is a light-receiving matrix; 551 is a CPU; and 561 and 562 are pattern presentation devices.

In this embodiment, the input part and feedback part of the optical content addressable memory system shown in FIG. 4 are replaced with a CPU, thereby facilitating the interface with electronic circuits.
Moreover, by replacing the output section with an electronic circuit, a hybrid optical associative memory system is created with one less spatial light modulation tube.

Note that the operating procedure is exactly the same as in the case of FIG. 4.

〔Effect of the invention〕

As described above, according to the present invention, compared to conventional associative memory devices that target electrical digital signals developed as computer storage devices, the iH[ - It is not necessary, and as a result, processing time can be greatly shortened, and output can be obtained continuously, making it possible to use it in optical search devices and the like.

In addition, in the case of associative memory, the memory contents are distributed throughout the memory, so even if a part of it is damaged, reading is not completely impossible.Also, by performing associative memory optically, it is possible to perform the same operation. It is possible to arrange a large number of circuits and easily interconnect them.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an optical content addressable memory device according to the present invention, and FIGS. 2, 3, 4, and 5 are optical systems of the optical content addressable memory device of the present invention using spatial light modulation tubes. FIG. 6 is a diagram for explaining the configuration of a spatial light modulation tube, which is a basic component of the optical content addressable memory device according to the present invention, and FIG. 7 is a diagram showing a crystal of the spatial light modulation tube. A diagram showing the secondary electron emission characteristics of the surface, Figures 8 and 9 are diagrams to explain the AND operation using the real-time hard clip function of the spatial modulation tube, and Figure 10 is a diagram to explain the enlarged imaging system. Figure 1
FIG. 1 is a diagram for explaining a multiplex imaging system, and FIG. 12 is a diagram for explaining an inverse multiplexing system. 1... Image, 2... Lens, 3... Photocathode, 4...
... Microchannel plate, 5 ... Mesh electrode, 6 ... Crystal, 61 ... Charge storage surface, 7 ... Half mirror 18 ... Monochromatic light, 9 ...; Analyzer, 10.
. . . Output light, 101 . . . Enlargement imaging system, 102 . . . Multiple imaging system, 103 . ... Inverse multiplex imaging system, 107 ... Threshold memory, 201 to 204
...Spatial light modulation tube, 211...Enlarged imaging system, 212
...Multiple imaging system, 213... Inverse multiple imaging system, 221
~228...Half mirror 1231~238...Reflector, Sl, S2, S3...Shutter. Applicant Hamamatsu Photonics Co., Ltd. Representative
Person Patent Attorney Masaru Hiru Kawa Nobuo Figure 1 Input 2 Nito y - Reminiscence Peak Figure 6 Figure 7 Figure 2 Figure i Encircle ■

Claims (1)

[Claims] an optical device for presenting an input or feedback signal; an enlarged imaging optical system for forming an enlarged image of the output from the optical device on a photocathode of a first spatial light modulation tube; a multiplex imaging optical system for forming multiple images on the photocathode of the modulation tube; a multiplication calculation output executed by sequentially reading out the first spatial light modulation tube and the second spatial light modulation tube;
Alternatively, an imaging optical system that images the output of the second spatial light modulation tube on the photocathode of the third spatial light modulation tube, and read out the second spatial light modulation tube and the third spatial light modulation tube in this order. an inverse multiplex imaging optical system that inversely multiplexes and images the output of the multiplication performed by Optical content addressable memory device.