RADER, Circuit Judge.
On summary judgment, the United States District Court for the Southern District of Texas held AllVoice Computing PLC’s (“AllVoice’s”) U.S. Patent No. 5,799,273 (“the '273 patent”) invalid because claims 60, 61, and 67 were indefinite under 35 U.S.C. § 112 ¶ 2 and because the patent specification did not disclose the best mode of practicing claim 73.
AllVoice Computing PLC v. Nuance Commc’ns, Inc.,
H-02-4471 (S.D.Tex. Feb. 22, 2006). Thus, the district court did not decide whether Nuance Communications, Inc.’s (“Nuance’s”) Dragon Naturally Speaking software infringes the '273 patent on voice recognition technology. Because the district court erred in applying both indefiniteness and best mode, this court reverses and remands.
I
The '273 patent covers an interface between a speech recognition engine and various end-user application programs on a personal computer. Claim 60, one of the asserted claims, reads:
60. A universal speech-recognition interface that enables operative coupling of a speech-recognition engine to at least any one of a plurality of different computer-related applications, the universal speech-recognition interface comprising:
input means for receiving speech-recognition data including recognised words;
output means for outputting the re-cognised words into at least any one of the plurality of different computer-related applications to allow processing of the recognised words as input text; and
audio playback means for playing audio data associated with the recog-nised words.
'273 Patent eol.29 11.22-34. When a user speaks into a computer’s audio input device, the computer’s speech recognition engine receives the message. At that point, the claimed invention creates an interface that facilitates translation of the apprehended message directly into application programs, typically word processors. Additionally, the interface in dependent Claim 61 maintains its own data structures • to keep track of the relative positions of words in the application programs. These data structures also preserve associations between those words and the original recorded speech. Claim 61 reads as follows:
61. The universal speech-recognition interface of claim 60, further comprising: means,
independent
of the one comput
er-related application, for forming link data linking a portion of the audio data to at least one the recognised words
independently
of the one computer-related application, the link data comprising:
one or more audio identifiers which link a portion of the audio data to one or more recognised words; and one or more position identifiers which link the recognised words to corresponding positions within the one computer-related application; and means,
independent
of the one computer-related application, for updating the position identifiers in response to changes in positions of the recognised words within the one computer-related application.
'273 Patent col.29 11.35-51 (emphases added). The interface in Claim 60 allows users to edit documents produced through speech recognition without losing the ability to play back the recorded speech in a coherent manner. Further, Claim 61 specifies the independence of the interface from any computer-related application. Claim 67, another independent claim, also includes the editing function and independence from any computer-related application.
Claim 73, held invalid for failing to satisfy the best mode requirement, includes instructions for performing a variety of functions:
73. A computer usable medium having computer readable instructions stored therein for causing a processor in a data processing apparatus to process recognition signals defining a string of recog-nised words and corresponding audio data to display the words and selectively play the audio data, the instructions comprising instructions for:
a) causing the processor to input the recognition signals from a speech recognition engine and the audio data, the recognition signals including a string of recognised words and audio identifiers identifying audio components corresponding to each recognised word;
b) causing the processor to implement an interface application program to receive the input recognised words and to input the recognised words into a processing application program to process the recognised words to cause the recognised words to be relatively moved;
c) causing the processor to implement the interface application program to
form link data linking the audio data to the recognised words,
said link data comprising the audio identifiers and information identifying the corresponding recognised words;
d) causing the processor to generate an image of the recognised words on a display;
e) causing the processor to receive a selection signal generated by a user for selectively identifying a word in the displayed words;
f) causing the processor to implement the interface application program to compare the identity of the selected word with said link data to identify any corresponding audio component; and
g) causing the processor to send the identified corresponding audio component to an audio playback device.
'273 Patent col.31 11.29-60 (emphasis added). The invention in claim 73, essentially a data link, keeps track of word position
changes to identify audio corresponding with the selected text.
II
This court reviews a grant of summary judgment without deference.
Dynacore Holdings Corp. v. U.S. Philips Corp.,
363 F.3d 1263, 1273 (Fed.Cir.2004). This court also reviews claim construction without deference.
Cybor Corp. v. FAS Techs., Inc.,
138 F.3d 1448, 1454 (Fed.Cir.1998) (en banc). “The review of indefiniteness under 35 U.S.C. § 112, paragraph 2, proceeds as a question of law without def erence.”
SmithKline Beecham Corp. v. Apotex Corp.,
403 F.3d 1331, 1338 (Fed.Cir.2005). A violation of the best mode requirement must be proved by clear and convincing evidence.
U.S. Gypsum Co. v. Nat’l Gypsum Co.,
74 F.3d 1209, 1212 (Fed.Cir.1996).
III
Through the claim construction performed by court-appointed expert, Professor Paul M. Janicke, the district court found that certain means-plus-function clauses in claims 60, 61, and 67 were indefinite. This court disagrees. All of these disputed clauses satisfy the claim definiteness requirement.
The definiteness requirement is set forth at 35 U.S.C. § 112 ¶ 2: “The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.” The test for definiteness asks whether one skilled in the art would understand the bounds of the claim when read in light of the specification.
Miles Labs., Inc. v. Shandon, Inc.,
997 F.2d 870, 875 (Fed.Cir.1993).
Before reviewing the bounds of the claim in light of the specification, the analysis requires attention to the level of skill assigned to a person of ordinary skill in the art. Unfortunately, the district court did not specify the proficiency of the hypothetical person of ordinary skill in the art that is essential to administering the definiteness test. During oral argument before this court, AllVoice’s counsel defined “a person of ordinary skill in the art” in the context of this case as “someone who has a degree in computer science or something equivalent and 2-3 years experience programming in Windows.” This definition is consistent with the level of skill ascertained in other software patent disputes.
See, e.g., Data Race, Inc. v. Lucent Techs., Inc.,
73 F.Supp.2d 698, 746 n. 330 (W.D.Tex.1999) (“Bachelor of Science degree in electrical engineering, computer science or 3-5 years of recent experience in the field”);
Katz v. AT & T Corp.,
63 F.Supp.2d 583, 594 n. 2 (E.D.Pa.1999) (“[a]t least a Bachelor’s degree in a scientific or engineering field, such as physics, electrical engineering, or computer science, and at least two years experience working in the field of computer telephony”). Because Nuance did not pose a different definition nor dispute the above definition, this court uses that characterization in applying the definiteness test.
The disputed clauses in claims 60, 61, and 67 qualify for means-plus-function treatment under 35 U.S.C. § 112 ¶ 6. “Claim construction of a means-plus-function limitation includes two steps. First, the court must determine the claimed function. Second, the court must identify the corresponding structure in the written description of the patent that performs the function.”
Applied Med. Res. Corp. v. U.S. Surgical Corp.,
448 F.3d 1324, 1332 (Fed.Cir.2006). “The determination of the claimed function and corresponding structure of a means-plus-function claim limitation is a question of law, reviewed de
novo.”
ACTV, Inc. v. Walt Disney Co.,
346 F.3d 1082, 1087 (Fed.Cir.2003). Under 35 U.S.C. § 112 ¶ 2 and ¶ 6, a means-plus-function clause is indefinite if a person of ordinary skill in the art would be unable to recognize the structure in the specification and associate it with the corresponding function in the claim.
Atmel Corp. v. Info. Storage Devices, Inc.,
198 F.3d 1374, 1381-82 (Fed.Cir.1999).
The clause at issue in claim 60 reads as follows: “output means for outputting the recognised words into at least any one of the plurality of different computer-related applications to allow processing of the recognised words as input text.” '273 Patent col.29 11.29-32. The district court, through the expert report of Professor Janicke — which it adopted in full— concluded that “this limitation requires a system outputting, alternately, to a plurality of different word processing or other application programs.”
AllVoice Computing PLC v. Scansoft, Inc.,
Civil Action No. H-02-4471 (S.D.Tex. Oct.13, 2005) (adopting in its entirety Mr. Janicke’s report and claim construction) (Janicke Report ¶ 34). The district court thus added the requirement that the system be able to output “alternately” to different programs. Although the output means must be capable of outputting recognized words to more than one program, the claim does not suggest the requirement that the means do so alternately. In other words, the district court incorrectly added the requirement that the user have the ability to change the destination program “on the fly.” A person skilled in the art would understand the clause to merely require compatibility with multiple output programs — not the ability of users to switch between multiple output programs at will.
In adding a limitation beyond the requirements of the claims, the district court emphasized a portion of the prosecution history beyond its true significance. During prosecution, AllVoice distinguished its invention from a program in the prior art called Digital Dictate. AllVoice noted that Digital Dictate had only a capability to work with a single application program. Janicke Report ¶ 34. This single application distinction, however, did not suggest that AllVoice’s invention must include the ability to work with multiple applications at once and change the destination program “on the fly.” At most, this distinction in the prosecution history merely requires that AllVoice’s patented interface work with multiple applications. This distinction did not mean that the interface must also output to different programs “alternately.” Thus, viewing the prosecution history in its proper context, this court construes the claim to require “a system capable of outputting to a plurality of different word processing or other application programs.”
With this clarification of claim meaning, the specification does contain structure corresponding to the “output means” clause of claim 60: “The speech recognition interface application 12 receives the recognised word at the head of the alternative list shown in FIG. 3 and outputs the word using the dynamic data exchange (“DDE”) protocol in the Windows operating system.” '273 Patent eol.7 11.3-7.
[[Image here]]
As explained in detail by AllVoice’s expert, Richard Sonnier, a person of ordinary skill in this art would understand the DDE protocol.
AllVoice Computing PLC v. Scansoft, Inc.,
Civil Action No. H-02-4471 (S.D.Tex. Aug. 11, 2004) (Second Supplemental Sonnier Decl. ¶ 14-15.). DDE is “a form of interprocess communication that uses shared memory to exchange data between applications.”
Id.
(referencing Microsoft Windows Software Development Kit Programmer’s Reference, Volume 1: Overview as Exhibit A.)
“A
skilled artisan reading the specification would recognize that numerous applications support the DDE transfer protocol and that preparing the software instruction to transfer recognized words (text) to third party wordpro-cessors, spreadsheets and other applications that typically receive text, would be a trivial matter well within the reach of a person of ordinary skill in the art.” Second Supplemental Sonnier Decl. ¶ 15. This court concludes that the reference to DDE in the specification is a structure corresponding to the “output means” clause of claim 60. With that understanding of the proper parameters of the claim, the record shows that an artisan of ordinary skill would understand the bounds of the claim when read in light of the specification.
KSR Int’l Co. v. Teleflex Inc.,
550 U.S. -, 127 S.Ct. 1727, 1742, 167 L.Ed.2d 705 (2007) (“A person of ordinary skill is also a person of ordinary creativity, not an automaton.”). Thus, the record shows that claim 60 satisfies the definiteness requirement.
The clauses at issue in claims 61 and 67 relate to the interface’s ability to perform functions “independent of’ the connected application program. Claim 61 has two contested clauses: “means, independent of the one computer-related application, for forming link data linking a portion of the audio data to at least one of the recognised words independently of the one computer-related application;” and “means, independent of the one computer-related application, for updating position identifiers in response to changes in positions of the recognised words within the one computer-related application.” '273 Patent col.29 11.38-41, 48-51. Claim 64 (on which claim 67 depends) has the clause: “means, independent of the computer-related application, for determining positions of the re-cognised words in the computer-related application.” '273 Patent col.30 11.7-9. The district court construed “independent of’ to mean essentially “isolated.” Janicke Report ¶ 56-60. According to this interpretation, the claimed interface cannot re
ceive any information back from the application. Rather, it must keep track of the positions of words in the application without ever obtaining position data from the application.
This interpretation of the claims is not correct. Instead, the claim term “independent of’ means that the interface must maintain its own position data, in its own data structures, but still have the ability to receive positional information from the application. Once again the district court went astray by taking the prosecution history out of context. As discussed earlier, AllVoice distinguished its invention during prosecution from an older software application program called Digital Dictate: “However, in contrast to the claimed invention, the ‘Microsoft Word’ version of Digital Dictate reviewed in the Seybold Reports
itself fails to include
any feature that determines the positions of recognized words. Instead of including these features
within its system,
Digital Dictate depends on the bookmarking feature inherent to Microsoft Word, the application program targeted to receive speech input.” Janicke Report ¶ 55 (emphases in original). The district court read this passage as a disavowal of any interface ability to receive information from the application program for use in maintaining position data.
To the contrary, this passage merely requires that the AllVoice interface include a feature that “determines [the] positions of the recognized words,” “within its own system.”
Id.
This distinction does not bar the interface from receiving any information from the application. As long as the interface maintains its own internal data structure to keep track of the positions of words, neither the prosecution history nor the claim language prevents the interface from using information from the application to maintain that data structure.
Thus, the district court proceeded to assess indefiniteness with an unduly narrow claim construction. This narrow reading of the “independent of’ requirement in claims 61 and 67 led to a finding that the specification did not contain structure corresponding to that narrow claimed function. With a proper reading of the claim term, however, the specification does contain adequate structure. Figure 2 and its associated description in columns 5, 6, and 7 of the patent show that memory maintains the position data separately from the application. Specifically, memory 20 stores the link data 25 (the position data stored by the interface 12) separately from the current document 26 maintained by the text processor 13. '273 Patent col.5 11.44-62.
[[Image here]]
Because neither the claim language nor the prosecution history requires more, the record shows that the district court erred in finding claims 61 and 67 indefinite on the basis of the “independent of’ language.
The district court also found claims 61 and 67 indefinite for failure to set forth sufficient algorithmic structure associated with the contested means-plus-function clauses. The district court held that the disclosure at col.7 1.7,
et seq.,
of the '273 patent, along with Figures 4 and 8A, did not constitute sufficient structure “for determining positions of the recognized words (claim 64 and indirectly claim 67) and for updating word positions after edits (claim 61)”. Janicke Report ¶ 60.
[[Image here]]
[[Image here]]
To the contrary, the specification contains sufficient algorithmic structure to give meaning to claims 61 and 67. Claim definiteness, as discussed earlier, depends on the skill level of a person of ordinary skill in the art.
Miles Labs., Inc.,
997 F.2d at 875. In software cases, therefore, algorithms in the specification need only disclose adequate defining structure to render the bounds of the claim understandable to one of ordinary skill in the art.
See, e.g., Med. Instrumentation and Diagnostics Corp. v. Elekta AB,
344 F.3d 1205, 1214 (Fed.Cir.2003) (“[H]ere there would be no need for a disclosure of the specific program code if software were linked to the converting function and one skilled in the art would know the kind of program to use.”)
See also Intel Corp. v. VIA Techs., Inc.,
319 F.3d 1357, 1366 (Fed. Cir.2003) (holding that the internal circuitry of an electronic device need not be disclosed in the specification if one of ordinary skill in the art would understand how to build and modify the device). In that connection, this record does not contain clear and convincing evidence that the disclosure at col.7 1.7, et
seq.,
of the '273 patent, along with Figures 4 and 8A of the patent, do not constitute sufficient structure to define the claim terms for the ordinarily skilled artisan.
To the contrary, the record contains the statement of Richard Sonnier. This statement set forth several straightforward ways that the algorithm represented in Figure 8A could be implemented by one skilled in the art using well-known features of the Windows operating system (messages, operating system function calls, and hooking). The Sonnier statement con-
eluded with the observation that “[a] person skilled in the art reading the '273 specification would know that any of these techniques could be used to determine the position of a recognized word in the third party application, would know the software to use and how to implement it.” Second Supplemental Sonnier Decl. ¶ 17. Thus, the record does contain sufficient algorithmic structure to give meaning to the claim terms from the vantage point of an ordinarily skilled artisan. Thus, Mr. Son-nier supplied the only assessment in this record of the adequacy of the specification to disclose enough steps to constitute an actual algorithm for carrying out the functions claimed in the means-plus-function clauses of claims 61 and 67. Without any record evidence to contradict Sonnier’s assessment, this court discerns that the district court erred in this indefiniteness judgment as well. Therefore, this court holds that claims 61 and 67 satisfy the definiteness requirement.
IV
The district court held the '273 patent invalid for deliberate concealment of the best mode — a requirement of 35 U.S.C. § 112 ¶ 1. A careful reading of the claim, however, shows that the alleged undisclosed best mode is not a best mode of practicing the claimed invention. To the contrary, the alleged best mode subject matter falls outside the scope of claim 73. Thus, the alleged best mode is not a way of practicing the claimed invention at all.
35 U.S.C. § 112 ¶ 1 provides that the specification of a patent “shall set forth the best mode contemplated by the inventor of carrying out his invention.” This requirement ensures a patent applicant discloses the preferred embodiment of his invention.
Teleflex, Inc. v. Ficosa N. Am. Corp.,
299 F.3d 1313, 1330 (Fed.Cir.2002) (“The purpose of the best mode requirement is to restrain inventors from applying for patents while at the same time concealing from the public preferred embodiments of the inventions they have in fact conceived.”).
Only the claimed invention is subject to the best mode requirement.
DeGeorge v. Bernier,
768 F.2d 1318, 1325 (Fed.Cir.1985) (reversing the Patent Appeals Board for extending the best mode beyond the proper claim scope);
Randomex, Inc. v. Scopus Corp.,
849 F.2d 585, 588, (Fed.Cir.1988) (“It is concealment of the best mode of practicing the claimed invention that section 112 ¶ 1 is designed to prohibit.”);
see Zygo Corp. v. Wyko Corp.,
79 F.3d 1563, 1567 (Fed.Cir.1996) (“[T]he parameters of a section 112 inquiry are set by the CLAIMS.”). According to § 112 ¶ 1, an inventor is required to disclose the best mode for “carrying out
his invention.”
(emphasis added) Because the claims represent “the subject matter which the applicant regards as his invention,” subject matter outside the scope of the claims also falls outside the scope of the best mode requirement. 35 U.S.C. § 112 ¶¶ 1-2;
see Engel Indus., Inc. v. Lockformer Co.,
946 F.2d 1528, 1531 (Fed.Cir. 1991) (“The best mode inquiry is directed to what the applicant regards as the invention, which in turn is measured by the claims.”);
Eli Lilly & Co. v. Barr Labs., Inc.,
251 F.3d 955, 963 (Fed.Cir.2001) (“[T]he extent of information that an inventor must disclose depends on the scope of the claimed invention.”).
To apply the best mode standard, a court must first “determine[] whether, at the time the patent application was filed, the inventor had a best mode of practicing the claimed invention.”
United States Gypsum Co. v. Nat’l Gypsum Co.,
74 F.3d 1209, 1212 (Fed.Cir.1996). This determination turns on the inventor’s own subjective beliefs.
Id. (citing Chemcast
Corp. v. Arco Indus. Corp.,
913 F.2d 923, 928 (Fed.Cir.1990)). “[T]he second part of the analysis [asks] ... has the inventor ‘concealed’ his preferred mode from the ‘public’?”
Chemcast Corp.,
913 F.2d at 928.
Nuance alleged, and the district court agreed, that some functions of WordEx-press, one of AllVoice’s commercial products, constituted an undisclosed best mode for claim 73. In reaching this judgment, the district court assumed that claim 73 required maintaining or updating link data during the editing process. Based on this assumption, the district court held that the best mode for maintaining links after text editing included the “Selrange macro” (“the macro”), Microsoft Windows “hooks,” and the disabling of certain features of Microsoft Word.
Claim 73 includes instructions causing a processor to perform a variety of functions. '273 Patent col.31 11.29-60. The instructions implement an interface program that allows audio data and recognized words to be input into a processing program, such as a word processor, in which the words are moved relative to each other.
Id.
The interface application also “form[s] link data linking the audio data to the recognised words....”
Id.
After a user selects a recognized word, the interface application uses the link data to identify audio corresponding with the selected word and plays back the audio corresponding to the selected word.
Id.
The macro monitors keystrokes to determine the changes to a document. Specifically, the macro detects a keystroke, determines whether the keystroke will add, delete, or move a character in the text, and determines the position of the cursor to ascertain the final effect of the keystroke.
Id.
The macro interacts with the “hooking” feature of the Microsoft Windows operating system to perform these functions. Because the macro cannot track the effects of non-keyboard editing techniques, the WordExpress program temporarily disables some features of the Microsoft Word word-processing program.
The district court incorrectly held that the macro’s functions were within the scope of claim 73. In light of the specification as a whole, this court construes “forming] link data” as the formation of the data structure in volatile memory, excluding operations to update or maintain link accuracy. '273 Patent col.7 1.1. The claim language dictates this meaning. Indeed, the district court also construed this limitation to mean “forming] link information about the words and audio signals” without reference to any updating or maintenance of link data. Additionally, Nuance admits, through its expert, that claim 73 does not claim the functionality to update link data: “there is no specific requirement that the interface be capable of playing back audio after a user has made arbitrary edits using the keyboard.”
In addition to the claim language, the doctrine of claim differentiation assists in arriving at a correct claim meaning in this case. Indeed “claim differentiation takes on relevance in the context of a claim construction that would render additional, or different, language in another independent claim superfluous.”
Curtiss-Wright Flow Control Corp., v. Velan, Inc.,
438 F.3d 1374, 1381 (Fed.Cir.2006);
see Phillips v. AWH Corp.,
415 F.3d 1303, 1314 (Fed.Cir.2005) (“Differences among claims can also be a useful guide in understanding the meaning of particular claim terms.”). In the instant case, claim 1 includes “means for forming link data” and “means for updating said link data.” The claims thus distinguish between “forming” and “updating.” These separate terms define separate processes accomplished by separate means. If this court interpreted the
process of forming link data to include processes to update that data, the second limitation (means for updating link data) would be superfluous. Independent claims 1, 15, and 28, among others, explicitly require forming link data, monitoring changes to text, and updating link data as distinct actions accomplished through separate means. '273 Patent col.16 11.32 — 46, col.18 11.51-67, col.22 11.4-13. Thus, monitoring and updating functions are not part of the forming link data limitation. The specification and the claims consistently use the terms “forming,” “updating,” and “monitoring” to denote separate processes with different end results. '273 Patent col.8 1.65 to col.9 1.12 (indicating that forming and updating are separate steps).
Unlike claims 1, 15, 28, and others in the '273 patent, claim 73 does not include “updating” link data, “maintaining” link data after editing, or “monitoring” changes to text. '273 Patent col.16 11.32— 46, col.26 11.1-6, col.73 11.29-60. Because “forming link data” does not include monitoring changes or updating link data, a method to perform those unclaimed functions falls outside the scope of claim 73. The macro considered by the district court does nothing more than monitor the changes to a document, to eventually facilitate the updating of link data. Claim 73 does not include these features. Thus, the macro cannot be a best mode for claim 73. Because the alleged best mode does not fall within the scope of claim 73, this court need not consider whether the inventors of the '273 patent actually believed the macro was their best mode of practicing the invention or whether they deliberately concealed that subject matter.
Nuance suggests that claim 73 implicitly includes updating or maintaining link data. To the contrary, each claim need not include every feature of an invention. Rather, this court enforces a “presumption that each claim in a patent has a different scope.”
Versa Corp. v. Ag-Bag Int’l Ltd.,
392 F.3d 1325, 1330 (Fed.Cir.2004) (quoting
Comark Commc’ns, Inc. v. Harris Corp.,
156 F.3d 1182, 1187 (Fed. Cir.1998)). Thus every claim need not contain every feature taught in the specification.
See, e.g., Nazomi Commc’ns, Inc. v. Arm Holdings, PLC,
403 F.3d 1364, 1369 (Fed.Cir.2005) (claims may “embrace [] different subject matter than is illustrated in the specific embodiments in the specification.”).
Because the functionality of the alleged best mode falls outside the scope of claim 73, this court detects no violation of the best mode requirement with respect to claim 73. Accordingly this court reverses the district court’s holding of invalidity of the '273 patent.
Y
For the reasons set forth above, this court reverses both the district court’s determination that claims 60, 61, and 67 were indefinite and its grant of summary judgment of invalidity for failure to satisfy the best mode requirement with respect to claim 73. Accordingly this court remands for further proceedings consistent with this opinion.
REVERSED AND REMANDED.
COSTS
Each party shall bear its own costs.