JPH03100531A - Optical arithmetic unit - Google Patents

Abstract

PURPOSE:To surely execute the sum-of-products operation at high speed with respect to information inputted optically or electrically by executing the modulation with respect to each luminous flux emitted from a light emitting means, in accordance with desired input information. CONSTITUTION:A space modulation element 2 and polarizing plates 2A, 2B constitute a space optical modulating means. In a state that input information is inputted one-dimensionally or two-dimensionally to the spatial optical modulating means, when plural luminous fluxes being parallel to each other are made incident on the spatial optical modulation means, each luminous flux is modulated in accordance with the state of the spatial optical modulating means in the incident position, that is, an input information value in the incident position. The process corresponds to the product of a luminous flux group and input information. When the luminous flux group modulated in such a manner is received by a light receiving means 4, the output of the light receiving means 4 becomes the sum of intensity of the modulated luminous flux group, and an obtained result is the sum-of-products operation. In such a manner, with respect to information inputted optically or electrically, the sum-of-products operation can be surely executed at high speed.

Description

[Detailed description of the invention] [Industrial application fields] The present invention relates to an optical arithmetic device.

[Prior Art] Arithmetic processing is an important process in information processing in general, but conventionally it has been electrically processed exclusively using electronic circuits or electric circuits.

[Problems to be solved by the invention] However, in recent years, in fields such as image processing and neural networks,
Proposals are being made to perform arithmetic processing using light.

The present invention aims to provide a novel optical processing device from this point of view.

[Means to solve the problem] The following five aspects of the present invention will be explained.

The optical arithmetic device of the present invention includes [a light emitting device, a spatial light modulating device, a one-separation device, and a light receiving device”.

The "light emitting means" emits a plurality of mutually parallel light beams.

The "spatial light modulation means" modulates each light beam from one light emitting means according to desired input information.

Information may be input to the spatial light modulation means by light or electrically.

The "separation means" separates each light flux that has been modulated by the spatial light modulation means from the optical path incident on the spatial light modulation means.

The "light-receiving means" receives the light beam group separated by the one-separation means.

Note that ``if multiple light beams from the light emitting means are arranged in a line'', linear information calculation can be performed, and ``spatial light modulation can be performed by arranging a group of light beams in a two-dimensional array''. If the means can perform two-dimensional spatial modulation,
Can process 2D data.

[Function] When a plurality of mutually parallel light beams are incident on the spatial light modulation means in a state where input information is input to the spatial light modulation means one-dimensionally or two-dimensionally, each light beam is shifted to the incident position. It is modulated according to the state of the spatial light modulation means at the time, that is, the input information value at the incident position. That is, this step corresponds to the product of the luminous flux group and the input information. When a group of light beams modulated in this manner is received by a light receiving means, the output of the light receiving means becomes the sum of the intensities of the group of modulated light beams. Therefore, the result obtained is a sum of products operation. Therefore, with the optical calculation device of the present invention, it is possible to perform neighborhood processing that is performed in digital image processing. Further, it is possible to input input to the spatial light modulation means by a semiconductor device, and in such a case, the present invention can perform calculation and transmission for the output of the semiconductor device.

[Example] Hereinafter, description will be given based on specific examples.

In the embodiment shown in FIG. 1, reference numeral 1 indicates a light emitting means, 2 indicates a spatial light modulator, 2A and 2B indicate polarizing plates, 3 indicates a separating means, and 4 indicates a photosensor as a light receiving means. ing.

The light emitting means 1 includes a semiconductor laser 10 as a light source, a collimating lens I2, and a light splitting element 14.

When the semiconductor laser IO is emitted, the emitted diverging light beam is collimated by the collimating lens 12, passes through the polarizing plate 2A, and enters the light splitting element 14. The light splitting element 14 is an element such as a hologram element, and splits the incident parallel light beam into a plurality of mutually parallel light beams each having a circular cross section.

It is assumed here that these plurality of light beams are arranged in a two-dimensional array. These luminous fluxes are assigned a series of serial numbers 1 to N.
Add. Further, the intensity of the luminous flux is represented by F, and the intensity of the i-th luminous flux in the serial number is F, (i=1 to N). In this example, the intensity Fi is due to the Gaussian intensity distribution of the semiconductor laser 10.

Next, the spatial light modulator 2 includes a photoconductive layer 25, a dielectric reflective layer 27, and an electro-optic crystal 29, which are laminated as shown in the figure.
These are sandwiched between transparent electrode films 21 and 23, and a DC voltage is applied between the electrode films 21 and 23 from a DC power supply 20. This spatial modulation element 2 and polarizing plate 2
A and 2B constitute spatial light modulation means.

The separation means 3 is composed of a beam splitter 30 and a lens array 32.

Photosensor 4 has a single light-receiving surface.

Now, the N pieces of mutually parallel light beams emitted from the light emitting means 1 are transmitted through the beam splitter 30 of the separating means 3, and then transmitted through the electrode film 23 and the electro-optic crystal 29 of the spatial light modulation element 2. reflected by 27 and redirected to the beam splitter.
When the beam is reflected by the beam splitter 30, it enters the lens array 32.

The lens array 32 has lenses arranged in an array corresponding to each reflected light beam, and each reflected light beam is converted into a focused light beam by the corresponding lens and enters the photosensor 4 via the polarizing plate 2B.

A DC power source 2 is connected between the electrode films 21 and 23 of the spatial light modulator 2.
Since a DC voltage of 0 is applied, the plane of polarization of each light beam reflected by the reflection N27 and emitted from the spatial light modulation element 2 is rotated from the state when it is incident on the spatial light modulation element 2. Therefore, the attitude of polarizing plate 2B! l! J Adjust.

The light flux incident on the polarizing plate 2B is efficiently transmitted to the photo sensor 4.
Make it incident on .

In the state described above, no information has been input to the spatial light modulation element 2 yet.

Here, as shown in FIG. 1, input information is input from above the spatial light modulator 2 as a two-dimensional optical image. Then, the input information passes through the electrode film 21, and the photoconductive layer 25 becomes a conductor depending on the light intensity of the information.

In the portion where the photoconductive layer 25 is made into a conductor, the DC voltage of the power source 20 is applied to the electro-optic crystal 29 between the reflective layer 25 and the electrode film 23, so the photoconductive layer 25 is not made into a conductor in this portion. The electric field acting on the electro-optic crystal 29 is stronger than in other parts, and the degree of rotation of the plane of polarization differs from that in other parts.

The input information is given as a two-dimensional image pattern as described above, but the pattern is decomposed into "pixels" corresponding to the N light beams incident on this pattern, and each pixel is divided into pixels in the same way as the light beams. Number them serially from 1 to N. That is, the i-th luminous flux is incident on the i-th "pixel".

Then, the plane of polarization of each light beam will rotate according to the information in the corresponding pixel, and the intensity of each light beam that passes through the polarizing plate 2B will differ depending on the degree of modulation by the corresponding pixel. It's happening.

For example, if the degree of modulation by the first pixel is G, the intensity of the i-th light beam transmitted through the polarizing plate 2B is proportional to the product (F, -G,).

Then, the output of the photosensor corresponds to Σ(F, ·Gi), and the desired sum-of-products calculation has been performed. Therefore, arithmetic processing such as neighborhood processing can also be performed by such a product-sum calculation. Of course, the above summation using the Σ sign means that the summation is performed from 1 to N for the suffix i.

Note that since the light beam from the semiconductor laser 10 is nearly linearly polarized, the two polarizing plates can be omitted.

Further, the lens array 32 is not necessarily required.

However, when the lens array 32 is used, the light flux from the spatial light modulator can be effectively extracted to the photosensor side while reducing the amount of noise mixed therein.

FIG. 2 shows another embodiment. To avoid clutter, the same symbols as in Figure 1 have been used for items that are not likely to be confused with each other.

The difference from the embodiment shown in FIG. 1 is that one light emitting means IA and the lens array 31 of the separating means 3A.

The light emitting means IA is composed of a semiconductor laser array 11 and a variable focal length lens array 13.

The semiconductor laser array 11 is an array of N semiconductor lasers arranged in a two-dimensional array.

The variable focal length lens array 13 includes variable focal length lenses,
The array is arranged in correspondence with the arrangement of semiconductor lasers in the semiconductor laser array 11.

The light beams from the corresponding semiconductor lasers are each converted into parallel light beams. Therefore, N mutually parallel light beams can be obtained by the light emitting means IA.

In this case, each light beam is given a serial number from 1 to N as in the previous embodiment, and the intensity of the first light beam is F.

The output obtained from the photosensor 4 corresponds to Σ(Fl-Gl), as in the embodiment of FIG.

In the embodiment shown in FIG. 1, the intensity of the luminous flux incident on the spatial light modulator 2 is 1! t F + was determined by the Gaussian intensity distribution of the semiconductor laser 1, but in the embodiment shown in FIG.

It is also possible to set F=constant for the intensity of the light flux F+k j = 1 to N, or it is also possible to realize a desired intensity distribution F by combining the semiconductor lasers emitted by the semiconductor laser array 11. A product-sum operation that adds the product of the intensity distribution and the input information is also possible, and the degree of freedom in the calculation is large.

Furthermore, in the case of the embodiment shown in FIG.
By simultaneously changing the focal length of each variable focus lens in step 3 and displacing the entire array in the optical axis direction (in the vertical direction in Figure 2), it is possible to continuously change the beam diameter of the N beams. can.

The lens array 31 is also an array of variable focal length lenses, and as described above, the variable focal length lens array 1
When the diameter of the light beam is changed by the action of 3, the lens array 31 also changes its focal length and its position accordingly so that it can condense the light beam whose diameter has changed.

In this way, by changing the beam diameters of the plurality of light beams obtained from the light emitting means, it is possible to change the size of the neighborhood when performing the above-mentioned wandering process.

In the embodiment shown in FIG. 2, the lens array 31 may be omitted, and the polarizing plate 2A may be omitted by aligning the polarization directions of the semiconductor lasers in the semiconductor laser array.

In each of the embodiments shown in FIGS. 1 and 2, a control plate can be provided between the beam splitter and the photodiode. Here, the control plate is plate-shaped or film-shaped, and its transmittance is distributed two-dimensionally, and modulation is applied to transmitted light according to the transmittance distribution. When such a control plate is used, a product calculation can be performed using the transmittance distribution of the control plate as an element of the product. It is also possible to use a control plate that provides a phase change distribution to the transmitted light in addition to the transmittance distribution.

In the two embodiments described above, the input information providing one of the elements of the product is optically input.

However, the input of information is not limited to optical input; it is also possible to input information electrically.

FIG. 3 shows only the characteristic parts of such an embodiment. A spatial light modulation element designated by the reference numeral 2A is used in place of the spatial light modulation element 2 in FIGS. 1 and 2.

Reference numeral 50 is a semiconductor device for inputting information, and has a plurality of output terminals 51.

The spatial light modulator 2A is formed by sandwiching a laminated dielectric reflective layer 27 and an electro-optic crystal 29 between an electrode film 22 and a transparent electrode film 23. The electrode film 22 is divided into parts corresponding to the terminals 51 of the semiconductor device w50, and each divided part is insulated from each other. Each portion is connected to the output terminal 51 of the semiconductor device on a 1:1 basis.

Input information is performed by outputting voltage signals to a desired combination of output terminals 51. The electrode film 23 is grounded, and when the input is performed as described above, the part of the electrode film 22 connected to the output terminal from which the voltage signal is output and the electrode film 23
An electric field acts on the electro-optic crystal between the two parts, and the plane of polarization of the light beam incident on this part rotates, allowing the sum-of-products operation to be performed in the same way as in the embodiments of FIGS. 1 and 2.

In this case, if the light emitting means IA used in the embodiment of FIG. A product-sum calculation can be performed, and in this case, if an area sensor or a line sensor is used in place of the photosensor 4, the output of the semiconductor device can be converted into time-series information using light as a medium.

FIG. 4 shows one example of applied usage of the optical arithmetic device of the present invention. Reference numerals 41, 42, and 43 indicate the second
The arrangement of the components is slightly different from the embodiment shown in the figure except for the photosensor 4, and each has the same structure. A plurality of such constituent parts are arranged in succession as shown in the figure, and input information is input to the spatial light modulation element in the constituent part 41, and at this time, the light flux emitted through the polarizing plate 2B is
The spatial light modulation element of the component 42 is inputted as input information. Then repeat this process for each component.

The light beam emitted from the polarizing plate of the last component 43 is received by the photosensor 4. In this way, the product operation in the product-sum operation can be performed in as many stages as the number of constituent parts.

In the component 41, instead of the spatial light modulation element 2,
Using the two spatial light modulators shown in FIG. 3, the initial information input may be performed electrically by a semiconductor device. It is also possible to use a control plate in more than one component.

[Effect of the invention] As described above, according to the present invention, a novel optical arithmetic device can be provided.

Since this device has the above-described configuration, it is possible to reliably and quickly perform product-sum operations on optically or electrically input information.

In addition, in the embodiment, calculations on two-dimensional input information have been explained, but as mentioned earlier, calculations on one-dimensional input information are also possible, and to realize this, it is necessary to It is only necessary to make the pattern of the input information in the optical modulation means into a line shape.

In addition, in the spatial light modulation element explained in the example, spatial light modulation was performed using rotation of the polarization plane due to the electro-optic effect.
If a liquid crystal is used instead of an electro-optic crystal in the spatial light modulation elements 2 and 2A, and the light modulation is intensity modulation, the polarizing plate 2A,
The spatial light modulation means can be configured only with the spatial light modulation element without using 2B.

[Brief explanation of drawings]

Fig. 1 is a diagram showing one embodiment of the present invention, Fig. 2 is a diagram showing another embodiment, Fig. 3 is a diagram showing only the characteristic parts of another embodiment, and Fig. 4 is a diagram showing one embodiment. FIG. 3 is a diagram for explaining that multi-stage calculations can be performed by applying the optical calculation device of the invention. 1000 light emitting means, 203. Spatial light modulator. 3200 minute history

Claims (1)

[Scope of Claims] A light emitting means for emitting a plurality of mutually parallel light beams, a spatial light modulation means for modulating each light beam from the light emitting means according to desired input information, and a modulation effect by the spatial light modulation means. An optical arithmetic device comprising: a separating means for separating each of the received light beams from an incident optical path to the spatial light modulating means; and a light receiving means for receiving a group of light beams separated by the separating means.