CITE:  888 F.2d 1370
CMON:  November 1989
PLAIN: In re Hiroyuki Iwahashi, Yoshiki Nishioka and Mitsuhiro Hakaridani
DEFND: United States Patent and Trademark Office
COURT: Court of Appeals for the Federal Circuit
DATE:  November 7, 1989

HISTORY:
Examiner's final rejection of claim relating to auto-correlation unit for
use in pattern recognition was affirmed by the United States Patent and
Trademark Office, Board of Patent Appeals and Interferences, and appeal
was taken.  The Court of Appeals, Rich, Circuit Judge, held that the
claim as a whole defined apparatus in the form of a combination of
interrelated means, and was directed to statutory subject matter, even
though the apparatus operated according to an algorithm.  REVERSED.

This appeal is from the decision of the United States Patent and Trademark
Office (PTO) Board of Patent Appeals and Interferences (board), dated
May 24, 1988, adhered to on reconsideration, affirming the examiner's
final rejection of the single claim of applicants' patent application
serial No. 454,022, filed December 28, 1982, entitled "Auto-Correlation
Circuit for Use in Pattern Recognition."  The sole ground of rejection is
that the subject matter claimed is nonstatutory under 35 U.S.C. Section
101 because it is merely a mathematical algorithm.  We REVERSE.

The real party in interest, according to appellants' brief, is Sharp
Kabishiki Kaisha (Sharp Corporation).


SUMMARY:
  Mathematical algorithm applied to a specific process is patentable.


JUDGE: RICH, Circuit Judge
Before RICH and BISSELL, Circuit Judges, and NICHOLS, Senior Circuit
Judge (Judge Nichols heard oral argument but, due to illness, did not
participate in the decision.)


DECISION:

The opening sentence of the specification states: "This invention relates
to an auto-correlation unit for use in pattern recognition to obtain
auto-correlation coefficients as for stored signal samples." The embodiment
more particularly discussed as a species of pattern recognition is voice
recognition.  The prior art calculation of auto-correlation coefficients
is described as being based on a calculation formula involving a
multiplication step. The specification states the disadvantage to be
as follows:

  Those state-of-the-art units for calculation of the auto-correlation
  coefficients have the disadvantage of requiring expensive multipliers
  and also complicated circuitry. As a result the auto-correlation unit
  circuitry within the entire pattern recognition apparatus is
  proportionately large and auto-correlation calculation demands a
  greater amount of time during recognition.
  . . . .


The principal object of this invention is to provide an auto-correlation
unit for pattern recognition which evaluates auto-correlation coefficients
by means of a simple circuitry without the need for an expensive multiplier
as well as eliminating the above discussed disadvantages.

Underlying the auto-correlation unit claimed, is a plethora of mathematical
demonstration by which the applicants purport to show that the approximated
value of the desired coefficient can be obtained without multipliers
by obtaining the  square  of the sum of two of the factors in the
equation and calculating the auto-correlation coefficient therefrom
according to a stated formula. The specification concludes:

  As explained in the foregoing, this invention offers a highly cost
  effective auto-correlation unit for pattern recognition with simple
  circuitry without the need to use an expensive multiplier, but which
  has comparatively high accuracy and can, moreover, calculate
  auto-correlation coefficients at high speed.

Fig. 1 of the application drawings is described as "a block diagram
schematically showing an embodiment of this invention" and appears
as follows:

   [SEE FIG. 1 IN ORIGINAL]

Fig. 2 is described as "a block diagram showing in more detail the
embodiment of this invention" and appears as follows:

   [SEE FIG. 2 IN ORIGINAL]

We shall not attempt a description of the electronic circuitry shown by
these drawings beyond explaining, for the better understanding of the claim,
that the units designated "ROM" and "RAM" are, respectively, a read only
memory and a random access memory, terms well understood by those skilled
in the art. "CPU" is a central processing unit. In this case, the function
of the ROM, 2 or 15, which is a permanent information storage device, is
to deliver as output the  square  of a number fed to it as input.
It is the electronic equivalent of a table in which one can look up the
square of numbers over a desired range.

We next reproduce the claim on appeal and do so by presenting a copy of
the claim as it has been presented in the Solicitor's brief, to which we
have added the letters in brackets designating at [a] the preamble or
introductory clause and at [b] through [h] the several means-plus-function
and other elements of the combination of elements recited. Under the heading
"Drawings" (the Solicitor's heading was "Fig. 1") we have copied verbatim
the Solicitor's designations. Element [d], it will be noted, is not in
means-plus-function form but specifies a "read only memory" or ROM, as
the Solicitor says. Element [f] is an anomalous clause seemingly intended
to indicate what data are stored in the ROM. It is not clear that a means
for storing anything in the ROM is part of the disclosed "unit" since the
application indicates that the squared values are "previously memorized"
in the ROM. However, that has nothing to do with the sole question before
us which is whether the claim as a whole is, in the words of the Solicitor,
"directed to nonstatutory subject matter," so we shall not comment further
on element [f].

                       Claim                                 Drawings

[a] An auto-correlation unit for providing auto-
     correlation coefficients for use as feature
     parameters in pattern recognition for N
     pieces of sampled input values
     X[n] (n=0 to N-1), said unit comprising:

[b] means for extracting N pieces of sample             Not shown in Fig. 1;
     input values X[n] from a series of sample          analog to digital
     values in an input pattern expressed with          converter 11 in
     an accuracy of optional multi-bits;                Fig. 2.

[c] means for calculating the sum of the                Adder 1.
     sample values X[n] and X[n] - Z
     (t = 0 - P, P < N);

[d] a read only memory associated with said             ROM 2.
     means for calculating;

[e] means for feeding to said read only memory          Signal path
     the sum of the sampled input values as an          connecting adder 1
     address signal;                                    to ROM 2.

[f] means for storing in said read only memory          Internal structure
     the squared value of each sum, (X[n] +             of ROM 2 after being
     X[n] - Z)2                                         programmed to store
                                                        squared values.

[g] means for fetching and outputting the               Read pulse (not
     squared values of each such sum of the             shown) which
     sample input values from said read only            applied to ROM 2;
     memory when said memory is addressed by            in Fig. 2, signal
     the sum of the sample input values; and            f[1] or f[2],
                                                        applied to ROM 15.

[h] means responsive to the output (X [n] +             Calculating circuit
     X[n] - Z) 2                                         5.
     of said read only memory for providing
     an auto-correlation coefficient for use
     as a feature parameter according to the
     following formula:

  [SEE FORMULA IN ORIGINAL]


This is one more in the line of cases stemming from the Supreme Court
decision in Gottschalk v. Benson, 409 U.S. 63, 34 L. Ed. 2d 273, 93 S. Ct.
253 (1972), decided by our predecessor, the United States Court of Customs and
Patent Appeals. They include in chronological order, In re Chatfield, 545 F.2d
152 (CCPA 1976), cert. denied 434 U.S. 875, 98 S. Ct. 226, 54 L. Ed. 2d 155,
(1977), In re Freeman, 573 F.2d 1237 (CCPA 1978), In re Maucorps, 609 F.2d
481 (CCPA 1979), In re Walter, 618 F.2d 758 (CCPA 1980), In re Abele,
684 F.2d 902 (CCPA 1982), and In re Meyer, 688 F.2d 789 (CCPA 1982). 
The list is not exhaustive but representative.

Out of these cases came the Freeman-Walter test to determine whether a claim
defines nonstatutory subject matter. It was stated in Freeman as follows:

  Determination of whether a claim preempts nonstatutory subject matter as
  a whole, in the light of Benson, requires a two-step analysis. First, it
  must be determined whether the claim directly or indirectly recites an
  "algorithm" in the Benson sense of that term, for a claim which fails
  even to recite an algorithm clearly cannot wholly preempt an algorithm.
  Second, the claim must be further analyzed to ascertain whether in its
  entirety it wholly preempts that algorithm.

573 F.2d at 1245. The opinion next discusses the meaning of "algorithm"
quoting from Chatfield footnote 5:

  Over-concentration on the word "algorithm" alone, for example, may mislead.
  The Supreme Court carefully supplied a definition of the particular
  algorithm before it [in Benson], i.e., "[a] procedure for solving a given
  type of mathematical problem." The broader definition of algorithm is
  "a step-by-step procedure for solving a problem or accomplishing some
  end." Webster's New Collegiate Dictionary (1976).

  . . . It would be unnecessarily detrimental to our patent system to
  deny inventors patent protection on the sole ground that their
  contribution could be broadly termed an "algorithm". [Emphasis of
  "sole" original, otherwise ours.]

In footnote 8 of the Freeman opinion the court further said:

  The preferred definition of "algorithm" in the computer art is: "A fixed
  step-by-step procedure for accomplishing a given result; usually a
  simplified procedure for solving a complex problem, also a full statement
  of a finite number of steps." C. Sippl & C. Sippl, Computer Dictionary
  and Handbook (1972).

Id. at 1246 n. 8. Appellants state that the apparatus claimed "may properly
be characterized as a computer."

We note these discussions of the meaning of "algorithm" to take the mystery
out of the term and we point out once again that every step-by-step process,
be it electronic or chemical or mechanical, involves an algorithm in the
broad sense of the term. Since Section 101 expressly includes processes as
a category of inventions which may be patented and Section 100(b) further
defines the word "process" as meaning "process, art or method, and includes
a new use of a known process, machine, manufacture, composition of matter,
or material," it follows that it is no ground for holding a claim is directed
to nonstatutory subject matter to say it includes or is directed to an
algorithm. This is why the proscription against patenting has been limited
to mathematical algorithms and abstract mathematical
formulae which, like the laws of nature, are not patentable subject matter.

The above-listed line of CCPA cases held some claims statutory and other
claims nonstatutory, depending entirely on what they said. We have to do the
same here. Appellants cautiously admit that their claim "at least indirectly,
recites an algorithm in some manner", and thus meets the first part of the
Freeman-Walter test, but argue strenuously and convincingly that it does not
meet the second part of the test, relying, inter alia, on the following
statement in Walter (footnote omitted):

  Once a mathematical algorithm has been found, the claim as a whole
  must be further analyzed. If it appears that the mathematical algorithm
  is implemented in a specific manner to define structural relationships
  between the physical elements of the claim (in apparatus claims) or
  to refine or limit claim steps (in process claims), the claim being
  otherwise statutory, the claim passes muster under Section 101.

618 F.2d at 767 (footnote omitted). Though the claim starts out by saying
in clause [a] that it is a "unit", appellants prefer to characterize what
they claim as apparatus with specific structural limitations.  By the
Solicitor's own analysis of the claim in the column labeled "Drawings",
supra, we are constrained to agree. Appellants emphasize that they specify
a ROM in clause [d] to which is fed an input from an adder specified in
clause [c].  The Solicitor states that [c] and [d] are connected together
by a signal path.  Next are means in the form of disclosed electronic
circuitry which take from the ROM its output in the form of squares of
numbers supplied as ROM input and feed them to a calculating circuit [h].
The claim as a whole certainly defines apparatus in the form of a
combination of interrelated means and we cannot discern any logical
reason why it should not be deemed statutory subject matter as either a
machine or a manufacture as specified in Section 101. The fact that the
apparatus operates according to an algorithm does not make it nonstatutory.
See In re Abele, 684 F.2d at 906. See also the discussion of that case
in In re Grams, 888 F.2d 835 (Fed. Cir. 1989), slip opin. at 11-12. We
therefore hold that the claim is directed to statutory subject matter.

In the Solicitor's brief the summary of argument states that the claim
"encompasses any and every means for performing the functions recited
therein."  We point out that the claim is a combination of means all but
one of which is a means-plus function limitation, the one exception being
the ROM, clause [d], which is a specific piece of apparatus. The claim
is therefore subject to the limitation stated in 35 U.S.C. Section 112
paragraph 6 that each means-plus-function definition "shall be construed
to cover the corresponding structure, material, or acts described in the
specification and equivalents thereof." {1}  This provision precludes the
Solicitor's interpretation of the claim. The Solicitor's summary also
contends that since the claim should be interpreted as he does, we should
regard it as though it were a method claim. Since he is wrong on the
first score, he is wrong on the second.

The decision of the board is REVERSED.

FOOTNOTES:

{1} The accuracy of this statement may be questioned in view of a sentence
in the opinion in In re Sweet, 55 C.C.P.A. 1191, 393 F.2d 837, 841-42 (CCPA
1968), which reads: "[A] recitation of "means" for performing a function
is interpreted broadly to cover all means capable of performing the stated
function and is not limited to the particular structure which the
application may disclose." (Emphasis added.) This statement, considered in
a vacuum, is partly true and partly untrue. It must be read, however, in
light of the opinion as a whole. It should not be removed from its context.
The untrue part is the initial statement that the means clause is
interpreted to cover all means to perform the function. It should have
said it is interpreted to cover the means disclosed and all equivalents
thereof which perform the function. The immediately preceding two
paragraphs of the opinion show that the court actually was reading the
"means" clause "in the light of 35 U.S.C. 112 [last paragraph]", just as
we are doing here. The statute is set forth in note 5. The truth of the
emphasized portion of the above sentence is beyond question because the
"means" clause includes equivalents of the disclosed structure. Section
112 para. 6 cannot be ignored when a claim is before the PTO any more
than when it is before the courts in an issued patent.