Professional Documents
Culture Documents
[54] PROCESSING SYSTEM FOR PERFORMING 3,937,942 2/ 1976 Bromley et al. .................. ,. 235/181
MATRIX MULTIPLICATION 4,120,035 10/1978 Cases et a1. ..... .. .. 364/602
‘ 4,286,328 8/1981 Booker ........ .. 364/851
[75] Inventors: Richard S. Schlunt, Loma Linda; 4,365,310 12/1982 Green 364/322
Stephen W. Decker, Sunland, both of 4,747,069 5/1988 Grinberg et al. ............. .. 364/841 X
cahf' Primary Examiner-Steven Mottola
[73] Assignee: General Dynamics Pomona Division, Attorney, Agent, or Firm-Neil F. Martin
Pomona, Calif. [57] ABSTRACT
[21] Appl' No‘: 131’478 An optical processing system for performing a matrix
[22] Filed: Dec. 10, 1987 multiplication Mx where M is a matrix and x is a vector
_ involves two two-dimensional arrays of optically trans
[gl] (31C: """""""""""" " Goa}; missive elements with each element of the ?rst array
[ 2] ' ' ' """"""""""""""""" " ’ 364/837’ communicating light to the corresponding element of
, _ the second array. The optical transmittance of each row
[58] Fleld of Search """"""""""" " 364gssgi’gg3l71’ 831,316 of one of the arrays is controlled according to the values
' ’ of the vector x, and the optical transmittance of each
[56] References Cited element of the other array is controlled according to the
U S PATENT DOCUMENTS corresponding matrix value rn(i,j). The ?rst array is
' ' uniformly illuminated and the amount of light transmit
;Vinter ------------------------------- - ted through the two arrays is detected by a suitable
’ ’ erg" " array of photodetectors. The output of each photode
Emir;"""" " tector will be proportional to the sum of the products of
3’492’469 V1970 Sig/err‘: 364/841 the vector x(i) with each value of the matrix in the
3:588:486 6/1971 Rosen I: 235/194 cwespm‘ding column 0f the matrix?rfay
3,592,547 7/1971 Noble .... .. 356/71
3,872,293 3/1975 Green ................................ .. 235/181 10 Claims, 2 Drawing Sheets
US. Patent Jun. 27, 1989 Sheet 1 of2 4,843,587
10
18/\\ f
'20
(\22
FIG. 1
, ‘ 14
12/
INPUT
28
% | ————————————————— _....S
—-->- N
——>- N
-—> N
FIG. 3
——> N
26’\_/ __> N g \
(/l I
16
M3”? WWZ‘EW
, \ %
[52
K431
FIG. 4
US. Patent Jun. 27, 1989 Sheet 2 of 2 4,843,587
\22
20
4,843,587
1 2
element with the data entered in the corresponding
PROCESSING SYSTEM FOR PERFORMING element of the column. The output from the array is
MATRIX MULTIPLICATION detected by a suitable arrangement of photodetectors
and can be converted into electrical signals representa
BACKGROUND OF THE INVENTION 5 tive of the results of a matrix multiplication.
The present invention relates to an optical processing In U.S. Pat. No. 4,286,328 of Booker an optical sys
system capable of performing matrix calculations, such tem is used to perform the computation of the radar
as matrix multiplication. ambiguity function or similar mathematical computa
Image processing, spectrum analysis, signal analysis tions. The system includes a programmable mask for
and other areas of data processing require the transfor storing a set of values, and a light source is positioned to
mation of data via a linear operator which is repre illuminate the mask and have its intensity modulated as
sented in the form of a matrix. Mathematically, the a function of the sample values. The image re?ected
operator is denoted by from the array is directed onto a photosensor array to
yield the ambiguity or other function.
y=M X (1) 15 In U.S. patent application Ser. No. 565,052 ?led Dec.
23, 1983 and assigned to the same assignee as the present
where M is the matrix m(i,j), i=l,m. . . R andj=l. . . application, an optical matrix multiplier is described
S (i.e. a matrix with R rows and S columns) and x is an which has the capability of handling a linear multiplica
R element row or column vector x(i),having the values tion operation involving a matrix having bipolar values.
i=1, . . . R. The value m(i,]) represents the element in 20 A light source produces a plurality of light beams w ith
the ith row and the jth column of the matrix. The value the intensity of each beam representative of a predeter
y is represented by the series [y(l) . . . y(S)]where mined value of a known column vector. The beams
illuminate a mask having individual elements forming a
matrix having an extra row to generate an offset in the
25 matrix values to ensure that the matrix has only unipolar
values. The resultant output from the photodetectors
It can be seen that the generation of y(k) requires R can be adjusted to subtract the scalar constant gener
multiplications and R additions, and this process must ated by the additional row of the matrix.
be repeated S times for all values of the vector y to be Some of the known optical processing systems for
calculated. Thus ‘the linear transform of data becomes performing complicated mathematical computations
computationally intense for relatively large values of R are relatively large or complex, require a large number
and S and real time applications are limited. The multi of optical components, and are subject to errors unless
plication and addition process can, of course, be per perfectly aligned. In some cases, a photographic mask
formed by a computer but the mathematical manipula or ?lm is used to simulate the matrix with the result that
tions will then be carried out sequentially and may take the device is not programmable.
an unnecessarily large length of time or require a costly,
complex and relatively large computer system, which is SUMMARY OF THE INVENTION
not suitable where space or weight is limited. According to the present invention an improved opti
Optical processing systems have been used in the past cal processing system for performing matrix multiplica
for performing mathematical manipulations involving tion is provided, which includes two two-dimensional
matrices with large numbers of elements, since such matrix arrays of optically transmissive elements. One of
systems can perform a large number of parallel addi the arrays has the tranmissivity of elements in each of its
tions and multiplications simultaneously. rows or columns controlled in accordance with compo
Such systems typically involve the direction of one or nents of a vector x and the other array has the transmis
more beams of light through an optical mask of variable sivity of each of its elements controlled in accordance
- transmittance. The intensity of the light beam or beams with components of a matrix M. The arrays are dis
is adjusted according to a ?rst set of values, and the posed ‘with elements of each array in registration with
transmittance of predetermined areas of the mask is corresponding elements of the other array, and a ?rst
controlled according to a second set of values. Thus if one of the arrays is uniformly lit by a suitable light
the light beams incident on each area or element of the source. Thus the intensity of light transmitted through
mask are controlled in accordance with the vector x and each pair of aligned elements of the two arrays will
the transmittances of the respective elements of the depend on the product of the transmittance of the two
mask are controlled in accordance with the values of elements and will be proportional to the product of one
the matrix M, a solution for the equation (1) above can of the vector values with one of the matrix values.
be obtained by measuring the amount of light passing Thus if all the light directed through a particular
through the mask. This is done by suitably arranged column j of the matrix is directed onto a suitable photo
photodetectors. detector, the output of the photodetector will be pro
In U.S. Pat. No. 3,592,547 of Noble, for example, portional to
analog data to be transformed is entered on a column of
resolution elements by suitable control of the optical
transmissivity of the elements, and a matrix transform R , (3)
45
55
65
UNITED STATES PATENT AND TRADEMARK OFFICE
CERTIFICATE OF CORRECTION
PATENT N0. : 4,843,587
DATED : June 27, 1989
INVENTUMS) 2 Schlunt, et al.
It is certified that error appears in the above-identified patent and that said Letters Patent is hereby
corrected as shown below:
Attest: