PROMPTU SYSTEMS CORPORATION v. COMCAST CORPORATION

IN THE UNITED STATES DISTRICT COURT

FOR THE EASTERN DISTRICT OF PENNSYLVANIA

PROMPTU SYSTEMS CORPORATION : CIVIL ACTION

:

v. : No. 16-6516

:

COMCAST CORPORATION, et al. :

MEMORANDUM

Judge Juan R. Sánchez September 26, 2025

This patent infringement action is before this Court for additional claim construction on

remand from the United States Court of Appeals for the Federal Circuit. At issue are four claim

terms from the two remaining patents-in-suit: three from U.S. Patent No. 7,047,196 (the ʼ196

patent) and one from U.S. Patent No 7,260,538 (the ’538 patent). After a claim construction

hearing on August 26, 2024, the Court issues this Memorandum to construe the disputed claim

terms.

BACKGROUND

Plaintiff Promptu Systems Corporation owns the ʼ196 and ʼ538 patents, both of which

concern voice-recognition technology. Broadly speaking, the ʼ196 patent “relates to using remote

voice-recognition systems to deliver requested (cable or video) content in response to a user’s

speech request,” while the ʼ538 patent “relates to using remote voice-recognition systems to

control a user’s television set based on a user’s speech command.” Promptu Sys. Corp. v. Comcast

Corp., 92 F.4th 1372, 1375 (Fed. Cir. 2024).

In December 2016, Promptu filed this action against Comcast Corporation and Comcast

Cable Communications, LLC (collectively, “Comcast”), asserting claims for, among other things,

infringement of the ʼ196 and ʼ538 patents. The case was stayed in July 2018 after the Patent Trial

and Appeal Board granted Comcast’s petitions to institute inter partes review of all asserted claims

of both patents. The stay was lifted in September 2020, and litigation resumed. In May 2022, this

Court issued a claim construction order, setting forth the Court’s constructions of eight disputed

claim terms identified by the parties. The following month, the Court issued a memorandum

explaining the basis for its rulings.

Based on the Court’s claim construction ruling, Promptu stipulated to the entry of final

judgment in favor of Comcast on the infringement claims as to the ʼ196 and ʼ538 patents, and the

Court directed that judgment be entered as to those counts.1 Promptu then appealed, challenging

the Court’s construction of four claim terms: “back channel,” “multiplicity of received identified

speech channels,” and “speech recognition system coupled to a wireline node” from the ʼ196

patent, and “central/centralized processing station” from the ʼ538 patent. In February 2024, the

Federal Circuit issued an opinion vacating the entry of final judgment on the infringement claims,

reversing in part and affirming in part this Court’s claim constructions, and remanding the case for

further proceedings. Promptu Sys. Corp., 92 F.4th at 1383.

On remand, the parties submitted new proposed claim constructions for each of the

disputed claim terms. The Court held a claim construction hearing on August 26, 2024.

DISCUSSION

Claim construction requires determining how a person of ordinary skill in the relevant art

at the time of the invention “would understand a claim term ‘in the context of the entire patent,

including the specification.’” Grace Instrument Indus., LLC v. Chandler Instruments Co., 57 F.4th

1001, 1008 (Fed. Cir. 2023) (quoting Phillips v. AWH Corp., 415 F.3d 1303, 1313 (Fed. Cir. 2005)

1 With the parties’ agreement, the Court also dismissed with prejudice all remaining counts of

Promptu’s Amended Complaint, including a clam for infringement of a third reissued patent, as

well as claims for breach of contract, unfair competition, promissory estoppel, and unjust

enrichment. See Dkt. Nos. 311, 316. Those claims are no longer part of this case.

(en banc)). In construing claim language, courts begin with the claims themselves, which “provide

substantial guidance as to the meaning of particular claim terms.” Phillips, 415 F.3d at 1314; see

also Grace Instrument Indus., 57 F.4th at 1008. The claims “do not stand alone,” however, but

“must be read in view of the specification, of which they are a part.” Phillips, 415 F.3d at 1315

(citation omitted). The specification is generally “the single best guide to the meaning of a

disputed [claim] term” and is thus “the primary basis for construing the claims.” Id. (citations

omitted). In addition to consulting the specification, a court should consider the patent’s

prosecution history, which is part of the intrinsic record, if it is in evidence. Id. at 1317. Beyond

the intrinsic record, a court may also rely on extrinsic evidence, such as “expert and inventor

testimony, dictionaries, and learned treatises,” though such evidence is “less significant than the

intrinsic record in determining the legally operative meaning of claim language.” Id. (internal

quotation marks and citation omitted).

The “general rule” that the words of a claim are to be given their “ordinary and customary

meaning” is subject to two exceptions. Thorner v. Sony Comput. Ent. Am. LLC, 669 F.3d 1362,

1365 (Fed. Cir. 2012). If the patentee “‘clearly set[s] forth a definition of [a] disputed claim term’

other than its plain and ordinary meaning,” then the patentee’s lexicography controls. Id. at 1365-

66 (quoting CCS Fitness, Inc. v. Brunswick Corp., 288 F.3d 1359, 1366 (Fed. Cir. 2002)).

Similarly, if the patentee “disavows the full scope of a claim term either in the specification or

during prosecution,” then the excluded feature “is deemed to be outside the reach of the claims of

the patent.” Id. (quoting SciMed Life Sys., Inc. v. Advanced Cardiovascular Sys., Inc., 242 F.3d

1337, 1341 (Fed. Cir. 2001)).

While “claims must be read in view of the specification, . . . it is improper to read a

limitation from the specification into the claims.” Liebel-Flarsheim v. Medrad, Inc., 358 F.3d 898,

904 (Fed. Cir. 2004) (internal quotation marks and citation omitted); see also Promptu Sys. Corp.,

92 F.4th at 1379 (“[P]articular features recited in the specification merely as aspects of

embodiments, and not expressly or even implicitly identifying requirements of the invention, do

not narrow a claim term that is otherwise broader in its ordinary meaning.”). The Federal Circuit

has also “counsel[ed] against interpreting the claims in a way that would omit a disclosed

embodiment absent clear evidence to the contrary.” Apple Inc. v. Corephotonics, Ltd., 81 F.4th

1353, 1359 (Fed. Cir. 2023); see also Kauffman v. Microsoft Corp., 34 F.4th 1360, 1372 (Fed. Cir.

2022) (“A claim construction that excludes a preferred embodiment is rarely, if ever correct and

would require highly persuasive evidentiary support.” (quoting Epos Techs. Ltd. v. Pegasus Techs.

Ltd., 766 F.3d 1338, 1347 (Fed. Cir. 2014)).

A. ʼ196 Patent

The ʼ196 patent discloses “[a] method and system of speech recognition presented by a

back channel from multiple user sites within a network supporting cable television and/or video

delivery.” 196 patent, Abstract. The Federal Circuit explained the patent addresses two problems

not solved in the prior art. First, the patent addresses the inability of prior-art set-top cable boxes

to support sophisticated speech recognition technology—which “‘requires state of the art

processing software and hundreds of megabytes of RAM to support’ it”—by “performing speech

recognition at a powerful remote computer, instead of at a local set-top box.” Promptu Sys. Corp.,

92 F.4th at 1374 (citations omitted). Second, because performing speech recognition at a

centralized processing site “requires the efficient upstream transmission (‘from subscriber to cable

plant’), using ‘limited bandwidth,’ of speech commands and content from many users

simultaneously,” the patent uses “‘a back channel containing a multiplicity of identified speech

channels’ where the ‘back channel is from a multiplicity of user sites.’” Id. (citations omitted).

1. “back channel”

The term “back channel” appears as a standalone term in asserted independent claim 27 of

the ʼ196 patent, which recites:

27. A system supporting speech recognition in a network, said system

comprising:

a speech recognition system coupled to a wireline node in said network for

receiving a back channel from a multiplicity of user sites coupled to said

network, further comprising

a back channel receiver, for receiving said back channel to create a received

back channel;

a speech channel partitioner, for partitioning said received back channel into

a multiplicity of received identified speech channels; and

a processor network executing a program system comprised of program

steps residing in memory accessibly coupled to at least one computer in

said processor network;

wherein said program system is comprised of the program steps of:

processing said multiplicity of said received identified speech channels to

create a multiplicity of identified speech content;

responding to said identified speech content to create an identified speech

content response, for each of said multiplicity of said identified speech

contents; and

wherein said network supports at least one of the collection comprising:

cable television delivery to said multiplicity of user sites; and video

delivery to said multiplicity of user sites.

ʼ196 patent, col. 55, ll. 9-36.

This Court previously construed the term “back channel” as “[a] fixed band of frequencies

or time slot(s) for transmitting signals to a speech processing system or engine,” adopting

Comcast’s proposed construction.2 Dkt. No. 299 at 2; Dkt. No. 315 at 15-16. The Federal Circuit

reversed, finding this Court erred by limiting the back channel to a path defined by a fixed band

of frequencies or time slots. The appellate court found “nothing in the meaning of the quite general

term ‘channel’” to support such a limitation. Promptu Sys. Corp., 92 F.4th at 1378. Turning to

2 The Court rejected Promptu’s proposed construction of “back channel” as a “stream of

communications transmitted upstream from set-top boxes.” Dkt. No. 315 at 15-16.

the specification, the court found the discussion therein “indicates . . . breadth regarding techniques

for the back channel.” Id. The specification “recognizes the possibility of using different protocols

or formats, including conversion between these, along different ‘subsections’ of the path or route

from user site to speech recognition system.” Id. It also “recognizes use of the DOCSIS cable-

modem protocol,” which can, but need not, use fixed frequency bands or time slots. Id. And it

describes transmission options that do not involve—or at least do not necessarily involve—fixed

frequency bands or time slots.3 See id. The appellate court thus concluded “[u]se of a fixed

frequency band or time slots, in [the ʼ196 patent], is no more than exemplary, not required.” Id.

The Federal Circuit declined to resolve the parties’ dispute as to whether the term “channel”

refers to “a path in which a data stream flows” (as Comcast argued) or “the data stream that flows

in the path” (as Promptu previously maintained), noting it was not clear how that dispute should

be resolved—or whether it was even necessary to resolve it. Id. at 1379-80. The appellate court

observed, however, that the specification appeared to use the term both ways, remarking that the

patent’s references to “‘processing’ or ‘receiving’ a ‘channel’” tend to suggest that, at least in those

contexts, “‘channel’ . . . means the data stream, not the path (it being awkward to describe a path

as being processed or, perhaps, received).” Id. at 1379-80. As an example of this dual usage, the

court noted the term “received back channel” had been given “an unchallenged construction as

meaning ‘[d]ata received via the back channel’—in which the ‘data’ term seems to use ‘channel’

3 Specifically, the Federal Circuit noted that through incorporation of U.S. Patent Application No.

09/664,874 (the ʼ874 application), the patent describes “signal transmission along the back channel

on dynamically assigned and managed (i.e., not fixed) frequency bands or time slots.” Promptu

Sys. Corp., 92 F.4th at 1378. The appellate court also noted the patent describes packet-based

transmissions, which Comcast’s expert agreed “may, or may not, be sent upstream using defined

frequency bands, or ‘channels.’” Id. (citation omitted).

to refer to data while the ‘via the back channel’ phrase seems to use ‘channel to refer to the path.’”

Id. at 1380.

On remand, the parties now agree the term “back channel” refers to a path over which data

is sent from set-top boxes to a speech recognition system. Pl.’s Br. 5, Dkt. No. 350; Defs.’ Br. 2-

3, Dkt. No. 349. Their agreement on these points, however, does not fully resolve their dispute.

Comcast argues the term “back channel” should be construed to mean “a single upstream path for

transmitting data, where the path begins at set-top boxes and ends at the speech recognition

system.” Defs.’ Br. 2, Dkt. No. 349. Promptu maintains the term should be construed as “a path

for receiving, at a speech recognition system, data transmitted from set-top boxes.” Pl.’s Br. 4,

Dkt. No. 350. The intrinsic record supports aspects of each party’s proposed construction but does

not fully support either one. For the reasons explained below, the Court will construe “back

channel” to mean “a path for transmitting data from set-top boxes to a speech recognition system.”

The parties first disagree as to whether the term back channel is limited to a single path.

Comcast argues the specification supports this limitation, noting the patent describes the invention

as “unique in that the speech command which originates at the user site . . . is sent upstream via

the return path (often five to 40 MHz) in the cable system to a central speech recognition and

identification engine.” Id. col. 5 ll. 22-26 (emphasis added); see Defs.’ Br. 5, ECF No. 349.

Comcast maintains this statement “emphasizes it is the use of a singular ‘return path’ that makes

the system ‘unique.’”4 Defs.’ Br. 5, ECF No. 349. Comcast further argues the back channel can

4 Elsewhere, Comcast characterizes this same passage as describing “the central location aspect of

the speech recognition system . . . as making the system unique.” Hr’g Tr. 141, Dkt. No. 372; see

also Defs.’ Responsive Br. 10, Dkt. No. 354.

contain multiple identified speech channels that originate from multiple different user sites, but

these speech channels must be part of the same single back channel. Id. at 5-6.

As Promptu notes, however, the specification also describes embodiments in which the

back channel is not limited to a single path. For example, the specification describes embodiments

in which data from groups of set-top boxes linked through different interfaces travels through a

wireline physical transport and an uplink to a receiver at an augmented node. See ʼ196 patent col.

27 ll. 36-42; id. col. 29 ll. 23-56; id. figs. 21, 23, 24. According to the patent, “[u]plink 1204

communication includes a back channel.” Id. col. 30 l. 4. The receiver, in turn, “provides 1326 a

back channel to speech engine 1330,” id. col. 30 ll. 7-8, or “provides a back channel to speech

engine 1330 through interactions 1328 with switch 1390 delivering 1392 the back channel to

speech engine 1330,” id. col. 32 ll. 1-3.

While the figures included in the patent depict a single wireline physical transport, uplink,

receiver, and switch delivering a back channel to a speech engine, see id. figs. 21, 23, 24, the patent

explains “[t]his is done strictly to simplify the discussion and is not meant to imply any limitation

on the invention,” id. col. 29 l. 66 – col. 30 l. 3. The patent makes clear there may be multiple

transports, uplinks, receivers, and switches, such that set-top boxes linked through different

interfaces may be serviced by different wireline physical transports and uplinks and thus may send

data along different paths to the speech recognition system. See id. col. 29 ll. 31-38, 44-52, 62-65

(explaining that “[o]ne uplink stream 1204 may be received from interface 170, and a second

uplink stream, not shown in FIG 21 or FIG. 23, may be received from interface 174”).

After noting that “[u]plink 1204 communication includes a back channel,” the patent goes

on to state “[t]his back channel includes multiple identified speech channels from multiple user

sites (STBs).” ʼ196 patent col. 30 ll. 4-7. But this description does not limit the back channel to

a single path. On the contrary, as noted, the patent states there may be “multiple uplinks 1204 . . .

provided by receiver 1322” (or by “more than one receiver 1322”), id. col. 29 ll. 46-48, or “multiple

uplink physical transports 1204,” id. col. 29 ll. 51-52. These descriptions of multiple uplinks that

“include[] a back channel” are inconsistent with Comcast’s argument that the back channel is a

singular path.5 The Court therefore finds the term “back channel” is not limited to a single path.

See Apple, Inc., 81 F.4th at 1359 (noting claims should not be construed in a way that would omit

disclosed embodiments).

The parties’ dispute as to the “upstream” limitation in Comcast’s proposed construction

presents a closer question. The parties previously agreed, both in this Court and in the Federal

Circuit, that the back channel flows upstream.6 The Federal Circuit recognized the parties’

agreement on this point, noting the term “‘[b]ack’ undisputedly refers to the direction (upstream)

opposite the direction of the programming (downstream).” Promptu Sys. Corp., 92 F.4th at 1378.

5 Comcast also argues the Federal Circuit’s decision in the previous appeal supports construing

“back channel” as a single path, pointing to examples of the appellate court referring to the back

channel in the singular. See Promptu Sys. Corp., 92 F.4th at 1375 (describing the back channel as

embodying a “many-in-one arrangement” (emphasis added)); id. at 1378 (“The patent discusses

the back channel, which contains a multiplicity of speech channels . . . .” (emphasis added)); id.

(noting the patent “recognizes the possibility of using different protocols or formats, including

conversion between these, along different ‘subsections’ of the path or route from user site to speech

recognition system” (emphasis added)). The Court finds these passing references to be of limited

relevance as the issue of whether the term “back channel” refers to a single path was not before

the appellate court.

6 See, e.g., Pl.’s Opening Suppl. Claim Construction Br. 5, Dkt. No. 260 (arguing the term “back

channel” in the ʼ196 patent “refers to communications sent upstream from set-top boxes”); Defs.’

Resp. Pl.’s Am. Opening Suppl. Claim Construction Br. 11, Dkt. No. 264 (noting “the asserted

claims make clear that the ‘back channel’ flows upstream”); Corrected Principal Br. for Appellant

24, No. 22-1939 (Fed. Cir. Oct. 20, 2022) (“The claims of the ʼ196 patent use ‘back channel’ to

refer to an upstream communications pathway over a network between multiple users and the

speech recognition system.”); Resp. Br. for Appellees 17, No. 22-1939 (Fed. Cir. Dec. 21, 2022)

(“There is no dispute that a ‘back channel’ is for upstream communications.”).

Even now, Promptu does not dispute that back channel flows at least primarily upstream.7 Pl.’s

Responsive Br. 7, Dkt. No. 355 (“The parties agree that ‘back’ in ‘back channel’ generally refers

to the direction of transmission: from the user towards the content provider.”); Hr’g Tr. 89-90,

Dkt. No. 372 (agreeing the back channel “is certainly upstream” in that “[t]he speech is not going

downstream[;] . . . [i]t is going up, meaning the direction opposite of the content”). Instead, the

parties’ dispute centers on whether the back channel is a solely upstream path (as Comcast argues)

or may permit some downstream communication (as Promptu maintains).

Comcast argues the specification “makes clear that the ‘back channel’ is an ‘upstream’

channel” because every time it “references a ‘back channel,’ it requires the direction of the data

flow in the back channel to be towards the receiver in the speech recognition system, rather than

towards the set-top boxes at the other end of the network.” Defs.’ Br. 4, Dkt. No. 349. As noted,

Promptu does not dispute that the patent shows the back channel flowing upstream. Rather,

Promptu argues the patent does not limit the back channel to a path that is solely upstream because

it describes embodiments in which speech data is sent to a speech recognition system using a

bidirectional path.

The examples Promptu offers of a bidirectional back channel fall into three categories.

First, Promptu notes the specification “describes some back channel implementations where the

same physical path carries both downstream and upstream traffic.” Pl.’s Responsive Br. 8, Dkt.

7 Indeed, in its opening claim construction brief, Promptu did not take issue with Comcast’s

characterization of the back channel as an “upstream” path, instead arguing that “[n]othing in the

specification limits the ‘back channel’ to a ‘single’ path shared end-to-end by multiple set-top

boxes.” Pl.’s Br. 7, Dkt. No. 350. Only after Comcast argued the back channel is “an ‘upstream’

path only,” Defs.’ Br. 3-4, Dkt. No. 349, did Promptu object to the limitation that the back channel

be used solely for upstream traffic, Pls.’ Responsive Br. 7-8, Dkt. No. 355. As a result, much of

Promptu’s argument on this issue, and all of Comcast’s response thereto, was made at the claim

construction hearing and not in the briefs.

No. 355. These include, for example, the descriptions of the wireline physical transport discussed

above, which is often depicted as carrying both upstream and downstream traffic. See, e.g., ʼ196

patent col. 29 ll. 34-48.

Second, Promptu points to examples in which downstream (or bidirectional)

communication is used to send or facilitate the sending of speech data upstream. See id. col. 11 ll.

38-40 (“The set-top box 1100 may be used for both upstream and downstream communication for

the described speech command function.”); id. col. 11. ll. 31-37 (describing a set-top appliance

that is “capable of receiving and decoding data in the downstream path,” using “[d]ownstream

data” containing “upstream channel allocation information” to “organize and synchronize the

transmission of upstream data”); id. col. 11 ll. 18-22 (“If a commercial set-top box 1100 is used to

transmit the upstream speech data, then the upstream channel allocation and transmission protocol

are controlled by the bi-directional communication system, which is resident in the set-top box.”).

Promptu argues these examples “undermine[] any assertion that the upstream and downstream

communication path must be entirely different.” Hr’g Tr. 44, Dkt. No. 372; see also Pl.’s

Responsive Br. 8, Dkt. No. 355 (“A back channel is still a back channel even if it occasionally

permits communications in the downstream direction so long as it can be, and is, used to send data

in the upstream direction.”).

Third, Promptu argues certain protocols identified in the patent are examples of using a

“bidirectional communication path to send data to a speech recognition system.” Hr’g Tr. 45, Dkt.

No. 372. Specifically, Promptu notes the patent recognizes use of the DOCSIS cable modem

protocol, ʼ196 patent col 11 ll. 51-59, which Comcast’s own expert described as a method of

sending information upstream and downstream over the cable network, see Kerr Dep. 67-69, Dkt.

No. 355-6 (describing DOCSIS as relating to “a logical channel . . . which is bidirectional,” i.e.,

“upstream and downstream,” and agreeing DOCSIS is a method of sending information to and

from the subscriber over the cable network). Promptu also notes the patent permits use of various

custom upstream protocols, including telephony, ʼ196 patent col. 12 ll. 59-65, which Comcast’s

expert agreed permits two-way communication on a telephone line, see Kerr Dep. 26, Dkt. No.

355-6.

Responding to these arguments, Comcast denies that the first category—embodiments in

which the same physical path (or medium) carries both upstream and downstream traffic—is an

example of a bidirectional back channel. Although the parties agree the back channel is a path,

Comcast maintains the physical medium the data traverses (e.g., a wire) is not itself the path. Hr’g

Tr. 69, Dkt. No. 372. Rather, a path is a channel on the physical medium—“for example, a radio

frequency channel.” Id. Thus, while it is possible to have communications going in two directions

on a given fiber or piece of coaxial cable, such two-way communication is not a bidirectional path

because “[w]hat you can’t have is communications going in two directions using the same range

of frequencies, the same bandwidth.” Id. at 74-75. With respect to the wireline physical transport

example, Comcast agrees the transport allows for communications in two directions, but not on

the same channel within the transport, as the patent makes clear that the transport includes a

“separate uplink and downlink, and it’s the uplink that contains the back channel.” Id. at 73, 76.

At the claim construction hearing, Comcast did not address Promptu’s argument that

limiting the back channel to a solely upstream path would exclude embodiments describing the

use of downstream communication to send speech data upstream. Notably, however, in its

briefing, Comcast relied on one of the same examples cited by Promptu—the statement that “[i]f

a commercial set-top box 1100 is used to transmit the upstream speech data, then the upstream

channel allocation and transmission protocol are controlled by the bi-directional communication

system which is resident in the set-top box”—as supporting the conclusion that upstream

communications occur in a solely upstream channel or path. See Defs.’ Br. 4, Dkt. No. 349 (citing

this passage as demonstrating that, “[a]ccording to the invention, set-top boxes use bandwidth

reserved for upstream communications to send data back to the cable head end”).

As to the third category of examples, Comcast disputes Promptu’s claim that the DOCSIS

cable modem protocol is an example of a bidirectional communication path, noting that the

DOCSIS specification in effect at the filing date of the ʼ196 patent states that “a single channel is

simplex,” meaning one-way. Hr’g Tr. 74 (quoting Pl.’s Ex. 5 at 13, Dkt. No. 355-5). The Court

understands Comcast to be arguing that while DOCSIS is a method of sending information to and

from the subscriber on a cable network, the communications occur over different channels.

Comcast also maintains the custom upstream protocols discussed in the patent8 are “radio

frequency-type communication[s]” and are “techniques for sharing a particular channel in a

particular direction.” Id. at 78. Yet Promptu argues one of those protocols—telephony—is a

method of two-way communication. Id. at 45, 90. Comcast did not respond to this argument at

the claim construction hearing, but in discussing the “three kinds of communications in RF

communications”—simplex, half-duplex, and duplex—Comcast explained that whereas a simplex

channel is one-way, “[d]uplex is like a phone call where everybody can talk and listen at the same

time.” Id. at 74. And, as Promptu notes, Comcast’s expert also agreed that using a single dial-up

link, “you could go both directions, upstream and downstream, on the telephone line.” Kerr Dep.

8 The patent states that speech and data signals may be “received and decoded by custom return

path receivers using at least one of the following protocol options: FM or AM

modulation/demodulation, FDMA, TDMA, FSK, PSK, or QPSK digital

modulation/demodulation, Spread Spectrum modulation/demodulation, Telephony, cellular

return, or Wireless.” ʼ196 patent col. 12 ll. 59-65.

26, Dkt. No. 355-6. This example supports Promptu’s position that the patent describes

embodiments in which speech data is sent to a speech recognition system using a path that permits

two-way communication and is a sufficient basis on which to reject the argument that the back

channel is limited to a solely upstream path. See Apple Inc., 81 F.4th at 1359. Because the

construction the Court adopts conveys that data is being sent upstream from set-top boxes to a

speech recognition system, the Court finds it unnecessary to specify that the back channel is an

upstream path.

Finally, the parties dispute whether the back channel should be defined as a path that

“begins at set-top boxes.” Comcast argues the Federal Circuit endorsed such a limitation with its

statement that the ʼ196 patent “discusses the back channel, which contains a multiplicity of

identified speech channels, broadly, defined by beginning at a user site (or sites) and ending at a

speech processing system.” Promptu Sys. Corp., 92 F.4th at 1378 (emphasis added) (citing ʼ196

patent col. 50 ll. 62-66; id. col. 22 ll. 8-14). As support for this statement, the Federal Circuit, in

turn, relied on language in the claims and specification characterizing the back channel as being

“from a multiplicity of user sites.”9 ʼ196 patent col. 50 ll. 62-64 (emphasis added); id. col. 22 ll.

12-13 (emphasis added). Comcast maintains the appellate court’s description is dispositive.

9 Specifically, the Federal Circuit relied on language from claim 1 reciting “[a] method of using a

back channel containing a multiplicity of identified speech channels from a multiplicity of user

sites presented to a speech processing system at a wireline node in a network supporting at least

one of cable television delivery and video delivery . . . .” ʼ196 patent col. 50 ll. 62-66. The

appellate court also relied on language from the specification describing figure 10 as “depict[ing]

a flowchart of a method that uses a back channel containing a multiplicity of identified speech

channels for speech recognition at a wireline node in a network supporting video or cable television

delivery in accordance with the invention. The back channel is from a multiplicity of user sites

and is presented to a speech processing system at the wireline node in the network.” Id. col. 22 ll.

8-14.

Promptu, in contrast, argues the patent uses the term “back channel” differently in different

contexts. According to Promptu, “back channel” can refer to “the path from one set-top box to the

speech recognition system.” Pl.’s Responsive Br. 4, Dkt. No. 355. But because the paths from

different set-top boxes may not overlap until further upstream, “back channel” can also refer to

“the part of the path shared by multiple set top boxes.” See id. at 3-5. Promptu maintains that in

the context of claim 27, which is recited from the perspective of a speech recognition system, the

term “back channel” refers to “the portion of the converged paths from multiple set-top boxes that

are received at the speech recognition system.” Id. at 4; see also Pl.’s Br. 5, Dkt. No. 350 (“The

words of claim 27 make plain that the ‘back channel’ refers to the end of a path that terminates at

a speech recognition system . . . and thus carries data from multiple user sites (e.g., set-top

boxes).”). Based on this understanding, Promptu argues “back channel” should be construed as

“a path for receiving, at a speech recognition system, data transmitted from set-top boxes.”

The difficulty with Promptu’s proposed construction is that the Federal Circuit interpreted

claim language in claim 1 of the patent as indicating the back channel “begin[s] at a user site (or

sites).” Promptu Sys. Corp. 92 F.4th at 1378. Promptu does not address this portion of the

appellate court’s opinion or explain why this Court is at liberty to disregard it. And while Promptu

suggests its proposed construction is specific to claim 27, the only asserted claim in which “back

channel” appears as a standalone term, “[i]t is well-established . . . that claim terms are to be

construed consistently throughout a patent.” Phil-Insul Corp. v. Airlite Plastics Co., 854 F.3d

1344, 1359 (Fed. Cir. 2017); see also Rexnord Corp. v. Laitram Corp., 274 F.3d 1336, 1342 (Fed.

Cir. 2001) (“[A] claim term should be construed consistently with its appearance in other places

in the same claim or in other claims of the same patent.”).10 Promptu does not adequately explain

why a different construction would be warranted, especially given that claim 1 also recites a back

channel that is received at a speech recognition system. ʼ196 patent col. 50 l. 62 – col. 51 l. 2.

For these reasons, and consistent with the Federal Circuit’s opinion, the Court agrees the

back channel “begin[s] at a user site (or sites) and end[s] at a speech processing system.” Because

the Court does not adopt the other aspects of Comcast’s proposed construction, the Court will

construe “back channel” as “a path for transmitting data from set-top boxes to a speech recognition

system.”

2. “multiplicity of received identified speech channels”

The term “multiplicity of received identified speech channels” appears in several asserted

claims of the ʼ196 patent, including claim 27, reproduced above. Adopting Comcast’s proposed

construction, this Court previously construed the term to mean “[m]ultiple different channels

carrying speech data, with each channel assigned to a particular user site,” wherein the term

“channel” referred to “a single band of frequencies or a designated time slot for transmitting

signals.” Dkt. No. 299 at 2; Dkt. No. 315 at 17-18; see also Dkt. No. 258-1 at 44. The Federal

Circuit reversed, finding “no stronger reason to limit an identified speech channel to a single

frequency band or designated time slot than to limit the ‘back channel’” to a fixed band of

frequencies or time slots. Promptu Sys. Corp., 92 F.4th at 1380. The appellate court also rejected

the requirement that each channel be “assigned to a particular user site,” noting this “associated-

10 This principle applies even though the Federal Circuit addressed the term “back channel” in the

context of claim 1, which was cancelled as a result of the inter partes review proceedings. See

Phil-Insul Corp., 854 F.3d at 1359 (declining to reconstrue claim terms that were construed in the

context of another claim in the same patent, which was cancelled during reexamination); Phillips,

415 F.3d at 1314 (noting unasserted patent claims can be “valuable sources of enlightenment as to

the meaning of a claim term”).

user-site limitation” appears in several dependent claims and therefore must be “presumed absent

from the ‘multiplicity of received identified speech channels’ phrase that provides its antecedent

basis.” Id. at 1381-82 (citing ʼ196 patent col. 51 ll. 12-13).

On remand, Comcast proposes that “multiplicity of received identified speech channels”

be construed to mean, simply, “more than one received identified speech channel,” with “channel”

defined consistently with its proposed construction for “back channel”—i.e., as a “path for data.”

Defs.’ Br. 7, Dkt. No. 349. Promptu’s proposed construction is “data received by a speech

processing system or engine.” Pl.’s Responsive Br. 9, Dkt. No. 355.11

The central dispute is whether “multiplicity of received identified speech channels” refers

to data or a path. Comcast argues the parties’ agreement that the “back channel” is a path for data

is dispositive of this issue. Citing the rule that “claim terms are to be construed consistently

throughout a patent,” Comcast maintains that because the parties agree the back channel is a path,

the term “channel” within the phrase “multiplicity of received identified speech channels” must

also be construed as a path. Defs.’ Br. 7 (quoting Phil-Insul Corp., 854 F.3d at 1359), Dkt. No.

349. Comcast further argues that apart from the term “channel,” the only term within the longer

claim phrase that requires construction is “multiplicity,” which Comcast suggests should be

construed as “more than one.”

11 Promptu initially proposed that “multiplicity of received identified speech channels” be

construed as a “plurality of data received by a speech processing system or engine identified as

originating from specific user sites.” Pl.’s Br. 8, Dkt. No. 350. In response to Comcast’s opening

brief, however, Promptu revised its proposed construction to eliminate “plurality of” from the

phrase “plurality of data” as data is already plural. Pl.’s Responsive Br. 9, Dkt. No. 355. Promptu

also eliminated the requirement that the data be “identified as originating from specific user sites”

as inconsistent with the Federal Circuit’s rejection of the “associated-user-site limitation” in this

Court’s previous construction. Id.

Promptu argues the claims make clear the term “multiplicity of received identified speech

channels” refers to data because they specify the “received back channel” is “partition[ed] . . . into

a multiplicity of received identified speech channels.” ʼ196 patent col. 51 ll. 3-4 (claim 1); id. col.

55 ll. 17-19 (claim 27). As the Federal Circuit noted, the term “received back channel” has “been

given an unchallenged construction as meaning ‘[d]ata received via the back channel.’” Promptu

Sys. Corp., 92 F.4th at 1380. Promptu maintains that because the “received back channel” is data,

“multiplicity of received identified speech channels” must also refer to data since data would be

partitioned into data, not a path, as both parties’ experts agree. Tinsman Decl. ¶ 51, Dkt. No. 355-

3 (“If data is partitioned, it is divided into data. . . . A path cannot be ‘partitioned’ into data, and

vice versa.”); Kerr Dep. 131, Dkt. No. 355-6 (agreeing data cannot be partitioned into a path). In

addition to reciting “partitioning [a] received back channel into a multiplicity of received identified

speech channels,” the claims recite that the “multiplicity of received identified speech channels”

are “process[ed] . . . to create a multiplicity of identified speech content.” ʼ196 patent col. 55 ll.

26-28; see also, e.g., id. col. 51 ll. 5-7 (claim 1); id. col. 53 ll. 1-3 (claim 14); id. col. 58 ll. 15-18

(claim 53). Promptu argues the patent’s references—in the claims and elsewhere in the

specification—to “processing a multiplicity of received identified speech channels” reinforce that

the “multiplicity of received identified speech channels” is data, not a path, based on the Federal

Circuit’s suggestion that the patent’s references to “‘processing’ or ‘receiving’ a ‘channel[]’ . . .

tend[] to suggest that ‘channel,’ at least in those references, means the data stream, not the path.”

Promptu Sys. Corp., 92 F.4th at 1380.

The Court agrees with Promptu that based on the claim language and settled construction

of “received back channel,” the term “multiplicity of received identified speech channels” refers

to data. Comcast argues Promptu’s assertion that data must be partitioned into data, not a path, is

unsupported. But, as noted, both parties’ experts agree with Promptu on this point. Tinsman

Decl. ¶ 51, Dkt. No. 355-3; Kerr Dep. 131, Dkt. No. 355-6. And while claim terms “are normally

used consistently throughout [a] patent,” Phillips, 415 F.3d at 1314, Comcast’s own constructions

reflect that “channel” is used differently in different contexts, treating “channel” as a path as used

in the term “back channel” but as data—or, more precisely, data received via a path—for purposes

of the term “received back channel.” As the Federal Circuit recognized, the now-settled

construction of “received back channel” as “data received via the back channel” appears to use

“channel” in two ways: “the ‘data’ term seems to use ‘channel’ to refer to data while the ‘via the

back channel’ phrase seems to use ‘channel’ to refer to the path.” Promptu Sys. Corp., 92 F.4th at

1380.

Under Promptu’s proposed construction, however, “multiplicity of received identified

speech channels” is only data—“data received by a speech processing system or engine.” Promptu

argues this construction follows the “same logic” as the construction of “received back channel,”

the other “received” channel term because that term uses “channel” to refer to data. Pl.’s

Responsive Br. 10, Dkt. No. 355. As noted, the Federal Circuit suggested otherwise, noting the

construction for “received back channel” uses “channel” to refer to both data and a path. Promptu

Sys. Corp., 92 F.4th at 1380. Comcast suggests that applying the logic of the construction of

“received back channel,” “multiplicity of received identified speech channels” should be construed

as “data received via a plurality of identified speech channels,” with “channel” referring to a path.

Defs.’ Responsive Br. 7, Dkt. No. 354; Hr’g Tr. 106, Dkt. No. 372.12 This construction is the only

12 The Court notes that Comcast’s own proposed construction for “speech recognition system”

treats “multiplicity of received identified speech channels” as “data received from identified

speech channels.” Defs.’ Br. 12, Dkt. No. 349.

one that is consistent with the construction of “received back channel.” It is also consistent with

the specification’s description of the back channel as “including” (or “containing”) a “multiplicity”

of (or “multiple”) identified speech channels. See ʼ196 patent col. 7 ll. 42-44; id. col. 22 ll. 8-10;

id. col. 30 ll. 5-6; id. col. 34 ll. 7-8; id. col. 50 ll. 62-63 (claim 1).

Promptu argues the patent “refers to ‘speech channels’ as data, not a path,” relying

primarily on two passages from the specification. Pl.’s Br. 9, Dkt. No. 350. The first describes

figure 10, which

depicts a flowchart of a method that uses a back channel containing a multiplicity

of identified speech channels for speech recognition at a wireline node in a network

supporting video or cable television delivery in accordance with the invention. The

back channel is from a multiplicity of user sites and is presented to a speech

processing system at the wireline node in the network. The speech processing

system performs the operations of the model.

ʼ196 patent col. 22 ll. 8-15. The second—a portion of the description of figure 23, which “depicts

a detail block diagram of an augmented distributor node 1310”—states:

Uplink 1204 communication includes a back channel. This back channel includes

multiple identified speech channels from multiple user sites (STBs) 1100, as shown

in FIGS. 21 and 22. Receiver 1322 provides 1326 a back channel to speech engine

1330. Speech engine 1330 performs at least the operations of FIG. 10.

Id. col. 30 ll. 4-9. Promptu asserts, without elaboration, that these passages “explain[] that the

‘multiplicity of received identified speech channels’ (from multiple user sites) are sent to the

speech engine over the back channel.” Pl.’s Br. 9, Dkt. No. 350. But nothing in either passage

equates speech channels with data. It seems more likely that a speech channel is a path, given the

parties’ agreement that the back channel that contains the speech channels is a path. Notably, in

discussing this “many-in-one arrangement,” the Federal Circuit observed that one method of

achieving it “‘may’ be through the partitioning scheme . . . set out in U.S. Patent Application No.

09/664,874,” which uses the term “channel” to refer to a path. Promptu Sys. Corp., 92 F.4th at

1374-75; see ʼ874 Application 4, Dkt. No. 259-24 (describing a method of assigning an

approximately equal number of active set-top boxes to the “available upstream data channels”).13

For these reasons, the Court construes “multiplicity of received identified speech channels”

to mean “data received via a plurality of identified speech channels,” whereas “channel” is to be

understood consistently with the term “back channel” as defined above.

3. “speech recognition system”

The term “speech recognition system” is part of the longer claim term “speech recognition

system coupled to a wireline node,” which appears in asserted claims 14, 27, and 53 of the ʼ196

patent. This Court previously construed “speech recognition system” to mean “a system whose

function is speech recognition,” “coupled to” as “connected in or near,” and “wireline node” as “a

network node providing video or cable television delivery to multiple users using a wireline

physical transport between those users at the node.” Dkt. No. 299 at 2; Dkt. No. 315 at 13-15. In

so holding, the Court adopted Comcast’s proposed constructions for “coupled to” and “wireline

node” and largely adopted Comcast’s proposed construction of “speech recognition system.” See

Dkt. No. 315 at 12-15.

On appeal, the Federal Circuit reversed as to each constituent term within the broader claim

phrase. The appellate court first found this Court’s “construction of ‘speech recognition system’

as ‘a system whose function is speech recognition’ improperly presupposes that speech recognition

is the exclusive function of the speech recognition system.” Promptu Sys. Corp., 92 F.4th at 1381

13 Promptu’s proposed construction also effectively reads the reference to “identified speech

channels” out of the claim term. As Promptu acknowledges, “[i]f data is partitioned, it is divided

. . . .” Pl.’s Responsive Br. 11, Dkt. No. 355. According to the claims, the data received via the

back channel is divided into multiple “received identified speech channels.” Yet under Promptu’s

proposed construction, the data, which is received at a speech recognition system, is divided into

“data received by a speech processing system or engine.”

(second emphasis added). The court noted the term itself “does not entail such exclusivity” and

that exclusivity is inconsistent with the specification, which describes functions other than speech

recognition that “a ‘[s]peech processor computer’ may perform,” such as billing and system

management. Id. The court declined to offer a “proper construction,” noting it was “not sure what

aspects of the phrase need clarification for resolution of the liability issues.” Id. Instead, the court

directed that “[a]dditional exploration should take place on remand, to the extent necessary, of the

best way to characterize the limitation of ‘system’ effected by the use of the phrasal modifier

‘speech recognition.’” Id.14 The court also cautioned further claim construction, “if needed at all,

should help resolve, not add to, uncertainty in the understanding the finder of fact is to use in

applying a claim term.” Id.

The Federal Circuit also rejected this Court’s construction of the term “coupled to” as

meaning “connected in or near,” finding the construction “incorrectly reads a proximity

requirement . . . into the claim term.” Id. Noting it had previously construed variants of the term

“coupled to” as meaning, simply, “connected to,” the appellate court found “no sufficient reason

to add a proximity requirement” in this case. Id. On the contrary, the court observed that while

the specification describes the invention as “comprising . . . a speech recognition system that is

centrally located in or near a wireline node,” it also states the speech recognition system may be

14 In this regard, the appellate court noted,

It may be relevant that the specification indicates that the slightly longer phrase,

“speech recognition processor system,” does not encompass the remote-control

device, the televisions, the telephone, the optical nodes, the set-top box, the gateway,

the server farm, or the server array—which are depicted as distinct elements separate

from the “speech recognition processor system.”

Id. (quoting ʼ196 patent fig. 3). The parties have not sought to clarify the “elements” of the speech

recognition system on remand.

centrally located in or near other sites, such as a server farm, website hosting facility, or network

gateway. Id. at 1382 (quoting ʼ196 patent col. 5 ll. 11-17). The court found the statement in the

specification “that a speech recognition system may be located ‘in or near’ sites other than the

wireline node further implies that proximity between the speech recognition system and the

wireline node (via coupling) is not required.” Id. The court instead “construe[d] ‘coupled to’ as

meaning ‘connected to,’” as it had done in other cases. Id.

Finally, as to the term “wireline node,” the Federal Circuit found this Court improperly

adopted the specification’s definition of “centralized wireline node,” even though the “centralized”

modifier did not appear in the claims. Id. The appellate court instead adopted Promptu’s proposed

construction of “wireline node” as meaning a “node connected by a wire, as opposed to a wireless

connection,” finding this construction to be “consistent with the scope of the claim language and

specification.” Id.

Because the Federal Circuit construed the terms “coupled to” and “wireline node,” only

the term “speech recognition system” remains for further construction on remand. Promptu argues

this term should be construed to mean a “system that performs speech recognition functions and

may also optionally perform other functions.” Pl.’s Br. 10, Dkt. No. 350. Comcast argues the

term should be construed to mean “[a] system in a central location directly connected to a node

connected by a wire, which at least processes data received from identified speech channels and

responds to identified speech content.” Defs.’ Br. 10, Dkt. No. 349.

The error the Federal Circuit identified in this Court’s previous construction of “speech

recognition system” was that it conveyed that speech recognition is the exclusive function of the

speech recognition system. Both proposed constructions attempt to correct this error, Promptu by

specifying the speech recognition system may “optionally perform” functions other than speech

recognition and Comcast by specifying that the speech recognition system must “at least” perform

two identified speech recognition functions without otherwise limiting the functions the system

may perform. Comcast also seeks to add other limitations, arguing the intrinsic record makes clear

that the speech recognition system must be “directly connected to a node connected to a wire” (i.e.,

to a wireline node) and “in a central location” vis-à-vis the multiplicity of user sites it serves. Id.

at 10-13 (emphasis added).

As to the first issue, Promptu does not dispute that the speech recognition system performs

“process[ing] data received from identified speech channels” and “respond[ing] to identified

speech content,” both of which are expressly recited in each of asserted claims. According to

Promptu’s own construction, the speech recognition system “performs speech recognition

functions,” and Promptu acknowledges the “processing” and “responding” steps are such

functions. See Pl.’s Responsive Br. 11-12, Dkt. No. 355; see also, e.g., ʼ196 patent fig. 10 & col.

22 ll. 8-15 (“speech processing system” performs the operations of the method depicted in figure

10, which include “processing multiplicity of received identified voice channels” and “responding

to identified speech content”); id. col. 30 ll. 8-9 (noting the speech engine “performs at least the

operations of FIG. 10”). Indeed, Promptu concedes “[t]here is very little daylight, if any, between

Promptu’s and this small aspect of Comcast’s proposal.” Pl.’s Responsive Br. 12, Dkt. No. 355.

Promptu instead argues its construction is preferable because it “more precisely conveys the

concept that the system must perform speech recognition functions and may perform other

functions” whereas Comcast’s construction “merely repeats other elements already expressly

recited in the claims.” Pl.’s Br. 11, Dkt. No. 350 at 11; see also Pl.’s Responsive Br. 12, Dkt. No.

355 (noting Promptu’s construction “us[es] the same words as the Federal Circuit”).

Comcast, however, argues its construction is necessary to make clear that it is the speech

recognition system that performs the “processing” and “responding” functions. Defs.’ Responsive

Br. 10, Dkt. No. 354. At the claim construction hearing, Comcast elaborated on the need for

clarification, explaining that claim language in certain of the claims that characterizes the

“processing” and “responding” steps as “program steps” of a “program system” may create

confusion for the jury as to whether the speech recognition system performs those functions. See

Hr’g Tr. 135, 145-46, Dkt. No. 372; see also, e.g., ʼ196 patent col. 52 l. 65–col. 53 l. 6 (describing

“processing” and “responding” as “program steps” of a “program system” that “control[s] at least

part of a speech recognition system”). Given this ambiguity, because the parties appear to agree

that the speech recognition system performs the “processing” and “responding” speech recognition

functions recited in the claims, and mindful that claim construction “should help resolve . . .

uncertainty in the understanding the finder of fact is to use in applying a claim term,” Promptu

Sys. Corp., 92 F.4th at 1381, the Court will adopt Comcast’s proposed construction on this issue.

As to the additional limitations Comcast proposes, as an initial matter, the Court agrees

with Promptu that Comcast’s argument that the speech recognition system must be “directly

connected to a node connected by a wire” runs afoul of the law of the case doctrine and the mandate

rule. The law of the case doctrine “posits that when a court decides upon a rule of law, that decision

should continue to govern the same issues in subsequent stages of the same case.” ArcelorMittal

France v. AK Steel Corp., 786 F.3d 885, 888 (Fed. Cir. 2015) (quoting Banks v. United States, 741

F.3d 1268, 1276 (Fed. Cir. 2014)). The doctrine encompasses the mandate rule, which “dictates

that ‘an inferior court has no power or authority to deviate from the mandate issued by an appellate

court.’” Id. (quoting Banks, 741 F.3d at 1268). Claim constructions established by the Federal

Circuit are law of the case on remand. Toro Co. v. White Consol. Indus., Inc., 383 F.3d 1326, 1336

(Fed. Cir. 2004); Ethicon Endo-Surgery, Inc. v. U.S. Surgical Corp., 149 F.3d 1309, 1315 (Fed.

Cir. 1998). As such, they are binding on the district court. ArcelorMittal France, 786 F.3d at 888

(holding law of the case doctrine prohibited a district court from revisiting claim construction

previously affirmed by the Federal Circuit); Epcon Gas Sys., Inc. v. Bauer Compressors, Inc., 90

F. App’x 540, 541 (Fed. Cir. 2004) (“The doctrine of law of the case requires lower courts to

follow the rulings of appellate courts on issues that have actually been decided by appellate

courts.”).15

In the previous appeal in this case, the Federal Circuit considered the meaning of “coupled

to” in the context of the broader claim phrase “speech recognition system coupled to a wireline

node” and “construe[d] ‘coupled to’ as meaning ‘connected to.’” Promptu Sys. Corp., 92 F.4th at

1382. The Federal Circuit did not remand the term for further construction or otherwise “invite

further refinements,” and its construction is now law of the case. AFG Indus., Inc. v. Cardinal IG

Co., 375 F.3d 1367, 1372 (Fed. Cir. 2004). While Comcast maintains it does not seek to challenge

or modify the Federal Circuit’s construction of “coupled to,” its proposed construction of “speech

recognition system” as requiring that the system be “directly connected” to a wireline node does

just that. Indeed, Comcast acknowledges as much, admitting its proposed construction “recognizes

that ‘coupled to’ in the context of the claims of the ʼ196 patent (and the Federal Circuit’s

15 A court may depart from the law of the case doctrine only in extraordinary circumstances, such

as “the discovery of new and material evidence not presented in the prior action, or ‘an intervening

change of controlling legal authority, or [a showing that] the prior decision is clearly incorrect and

its preservation would work a manifest injustice.’” Toro Co., 383 F.3d at 1336 (quoting Intergraph

Corp. v. Intel Corp., 253 F.3d 695, 698 (Fed. Cir. 2001)). But cf. W.L. Gore & Assocs., Inc. v.

Garlock, Inc., 842 F.2d 1275, 1279 (Fed. Cir. 1988) (holding a district court was bound by the

Federal Circuit’s interpretation of a claim term as law of the case on remand but recognizing the

Federal Circuit could “reconsider the decision if certain ‘exceptional circumstances’ exist”),

abrogated on other grounds as recognized by Zoltek Corp. v. United States, 672 F.3d 1309 (Fed.

Cir. 2012). Comcast does not argue extraordinary circumstances are present here.

construction) means ‘directly connected to.’” Defs.’ Br. 10-11, Dkt. No. 349 (emphasis added).

This attempt to revise the Federal Circuit’s construction of “coupled to” under the guise of another

claim term is improper under the law of the case doctrine and the mandate rule. See, e.g., AFG

Indus., 375 F.3d at 1372 (holding a district court erred by reevaluating on remand the construction

of the term “layer” adopted by the Federal Circuit on appeal).16

The Court also declines to adopt Comcast’s proposed limitation that the speech recognition

system must be “in a central location.” Promptu argues imposing such a limitation would violate

the mandate rule because the Federal Circuit rejected a location requirement for the speech

recognition system in the previous appeal. This Court disagrees. In reversing this Court’s prior

construction of “coupled to” as meaning “connected in or near,” the Federal Circuit rejected a

proximity requirement, via coupling, between the speech recognition system and the wireline node.

Promptu Sys. Corp., 92 F.4th 1382. Promptu suggests proximity is just another way of saying

location, see Hr’g Tr. 121, Dkt. No. 372, but the appellate court’s proximity discussion concerned

only the location of the speech recognition system relative to the wireline node. The appellate

court was not asked to consider whether the speech recognition system must be centrally located

in some broader sense. And while this issue does not appear to relate to the additional exploration

the Federal Circuit contemplated would occur on remand regarding “the best way to characterize

16 Comcast’s argument on this point is also unpersuasive because the claims themselves—and the

larger claim term presented for construction—already require that the speech recognition system

be “coupled to a wireline node”—i.e., “connected to” a “node connected by a wire, as opposed to

a wireless connection.” Construing the constituent term “speech recognition system” to mean a

system “directly connected to a node connected by a wire” would render the additional constituent

claim terms “coupled to” and “wireline node” redundant. See Apple, Inc. v. Ameranth, Inc., 842

F.3d 1229, 1237-38 (Fed. Cir. 2016) (rejecting claim constructions that included features of the

claim term “already recited in the claims” because they “would make those expressly recited

features redundant”).

the limitation of ‘system’ effected by the use of the phrasal modifier ‘speech recognition,’”

Promptu Sys. Corp., 92 F.4th at 1381, the Court does not understand the appellate court’s mandate

to preclude its resolution.17

As to the merits of the parties’ dispute, while the claims do not use the term “central

location” (or “centrally located”), Comcast nevertheless argues the patent requires the speech

recognition system to be in a central location because the specification “characterizes ‘the

invention’ as a centrally located [speech recognition system].” Defs.’ Responsive Br. 10, Dkt. No.

354. The Federal Circuit has held that “[w]hen a patent . . . describes the features of the ‘present

invention’ as a whole, this description limits the scope of the invention.” Verizon Servs. Corp. v.

Vonage Holdings Corp., 503 F.3d 1295, 1308 (Fed. Cir. 2007); see also C.R. Bard, Inc. v. U.S.

Surgical Corp., 388 F.3d 858, 864 (Fed. Cir. 2004) (noting statements in the specification “that

describe the invention as a whole, rather than statements that describe only preferred embodiments,

are more likely to support a limiting definition of a claim term”). Comcast argues this rule applies

here, pointing to the specification’s “Summary of the Invention” section, which describes “[t]he

invention” as “compris[ing] a multi-user control system for audio visual devices that incorporates

a speech recognition system that is centrally located in or near a wireline node . . . .” ʼ196 patent

col. 5 ll. 11-13 (emphasis added); see also id. col. 5 ll. 22-26 (“This system is unique in that the

17 Promptu also argues the “central location” limitation is improper because the Federal Circuit

held this Court erred in adopting the specification’s definition of “centralized wireline node” as

the construction for the unmodified claim term “wireline node.” In reversing this Court’s

construction, the Federal Circuit noted “[t]he absence of the ‘centralized’ modifier in the claims

. . . means that the claimed ‘wireline node’ must be broader in some way than the ‘centralized

wireline node’ defined in the specification.” Promptu Sys. Corp., 92 F.4th at 1382. This ruling is

inapposite to the present dispute as the specification contains no lexicography for a “centrally

located speech recognition system.”

speech command which originates at the user site . . . is sent upstream via the return path . . . in

the cable system to a central speech recognition and identification engine.” (emphasis added)).

The specification goes on to state, in the sentence immediately following the one cited by

Comcast, that “[t]he speech recognition system may also be centrally located in or near a server

farm[,] a web-site hosting facility, or a network gateway.” Id. col. 5 ll. 15-17. The passage as a

whole thus makes clear the speech recognition system is not limited to a particular location. See

Promptu Sys. Corp., 92 F.4th at 1382 (recognizing the speech recognition system may be located

‘in or near’ sites other than the wireline node”). And even assuming that the descriptions of the

speech recognition system as “centrally located” are limiting, the Federal Circuit, addressing this

same passage in the specification, equated the “centrally located” descriptor to “performing speech

recognition at a powerful remote computer, instead of at the local set-top box.” Id. at 1374

(discussing ʼ196 patent col. 5 ll. 11-17).18 It is unnecessary to add the requirement that the speech

recognition system be “in a central location” because the claims already make clear that the speech

recognition system is separate from the local set-top box (or “user site”). See, e.g., ʼ196 patent

col. 55 ll. 11-14.19

18 Specifically, the appellate court noted the patent addressed one problem not solved in the prior

art—namely, the inability of prior-art set-top boxes to support sophisticated speech recognition

technology—through “‘a speech recognition system that is centrally located in or near a wireline

node’ or is ‘centrally located in or near a server farm,’ ‘a web-site hosting facility,’ or ‘a network

gateway’—i.e., through performing speech recognition at a powerful remote computer instead of

at a local set-top box.” Promptu Sys. Corp., 92 F.4th at 1374.

19 Moreover, it is difficult to see how adding a “central location” limitation would help resolve

uncertainty in the jury’s understanding, given the absence of a reference point for the

centralization.

Accordingly, the Court construes “speech processing system” as a “system that at least

processes data received from identified speech channels and responds to identified speech

content.”

B. ʼ538 Patent

The ʼ538 patent discloses a “method and apparatus . . . for remotely processing voice

commands for controlling a television.” ʼ538 patent col. 2 ll. 22-23. The specification notes that

“[a] problem with the prior art voice recognition systems is that they require a sophisticated voice

recognition system in close proximity to the user, requiring individual units which is quite costly.”

Id. col. 1 ll. 59-62. The patent addresses this problem by processing voice commands remotely,

at a “centralized voice command processing system” that “services a multitude of users.” Id. col.

1 ll. 62-63.

1. “central/centralized processing station”

The term “centralized processing station” (or “central processing station”) appears in

asserted claims 34 and 40 of the ʼ538 patent. Claim 34 claims

A centralized muti-user voice operated television control system, comprising:

television remote controls configured to directly and wirelessly control

television sets and additionally to receive user voice input and wirelessly

transmit first output representative of the voice input to television set-top

boxes,

television set top boxes configured to receive television input signals via cable

television link and provide television output signals compatible with

television sets, the set top boxes additionally responsive to receiving the

first output from the television remote controls to transmit representative

second output to a central processing station via the cable television link;

a centralized processing station configured to receive and process second

output from a multitude of television set top boxes by applying voice

recognition to the second output to identify user-intended voice commands,

to derive set-top-box-compatible instructions to carry out the identified

voice commands, and returning signals representing the instructions to

respective top boxes via the cable television link;

where the set top boxes are further responsive to receiving the signals

representing the instructions from the central processing station to execute

the instructions.

Id. col. 13 ll. 37-61 (emphasis added).

Adopting Comcast’s proposed construction, this Court previously construed the term

“centralized processing station” to mean a “[d]evice at a cable-TV network head-end unit that

receives and performs voice recognition on voice commands, and generates and returns

instructions to set-top boxes to carry out the commands.” Dkt. No. 301; Dkt. No. 315 at 7. On

appeal, Promptu challenged both the requirement that the centralized processing station be located

“at a cable-TV network head end unit” (the “location requirement”) and the requirement that the

centralized processing station itself must “perform[] voice recognition on voice commands” (the

“function requirement”). Promptu Sys. Corp., 92 F.4th at 1382.

The Federal Circuit affirmed as to the function requirement, agreeing “the claim language,

read as a whole, specifies that the ‘centralized processing station’ (or some element of it) performs

the recited speech recognition.” Id. As to the location requirement, however, the Federal Circuit

reversed, finding this Court “incorrectly required the recited ‘device’ to be located ‘at a cable-TV

head-end unit.’” Id. In so holding, the appellate court noted the patent makes clear that a “head-

end unit” and a “centralized processing station” are different things, as “the two phrases are

different claim elements used in separate claims throughout the ʼ538 patent.” Id. And while the

specification describes a centralized processing station located at a cable television head-end

unit—and, in one diagram, uses the same number to refer to both elements—the court found these

references to be merely exemplary. The court concluded:

Such a limitation [i.e., requiring the centralized processing station and the head end

unit to be to be colocated] would read embodiment features into a claim term that

is not so limited. The “centralized” term in “centralized processing station,” as the

claim language makes clear, refers simply to its relation to a group of set-top boxes.

See Comcast Response Br. at 61 (“The claim term ‘centralized processing station’

makes clear that the processing station is a discrete physical thing (a ‘station’) that

is centrally located (‘centralized’) vis-à-vis the ‘multitude of television set top

boxes’ from which it receives signals.” (citing ʼ538 patent, col. 13, lines 37-61)).

Such centralization exists when the station is in the cable network, serving many

set-top boxes, whether or not it is located at the same place as a head-end unit.

Id.

Consistent with the Federal Circuit’s ruling on the function requirement, the parties now

agree the centralized processing station is a device that “receives voice commands, performs voice

recognition on voice commands, and generates and returns instructions to set-top boxes to carry

out the commands.”20 Defs.’ Br. 13, Dkt. No. 349; Pl.’s Br. 14, Dkt. No. 350. They continue to

disagree, however, as to the meaning of “centralized.” In their written submissions to the Court,

the parties agree the word “centralized” means “centrally located,” but disagree as to the

appropriate reference point. Comcast argues the centralized processing station must be “centrally

located in the cable network with respect to the set-top boxes it serves.” Defs.’ Br. 13, Dkt. No.

349 (emphasis added). Promptu argues the processing station need only be “centrally located

relative to multiple set top boxes.” Pl.’s Br. 14, Dkt. No. 350 (emphasis added). At the claim

construction hearing, Promptu alternatively argued the Court need not construe the term

“centralized procession station” further but should instead simply excise the location requirement

from this Court’s previous construction of the term. Hr’g Tr. 155, Dkt. No. 372. Promptu

maintains both of its proposed constructions should be regarded as law of the case—either on the

20 The parties’ agreed-upon phrasing differs only minimally from the language considered by the

Federal Circuit. Compare Promptu Sys. Corp., 92 F.4th at 1382 (describing a device “that receives

and performs voice recognition on voice commands, and generates and returns instructions to set-

top boxes to carry out the commands” (emphasis added)), with Dkt. No. 348 at 2 (describing a

device “that receives voice commands, performs voice recognition on voice commands, and

generates and returns instructions to set-top boxes to carry out the commands” (emphasis added)).

theory that the Federal construed the term “centralized” as Promptu proposes or because the

Federal Circuit reversed the location requirement in this Court’s previous construction but did not

indicate it was remanding the claim term for further construction. See id. at 154-56.

Although it is not clear that the law of the case doctrine applies here, the Court agrees with

Promptu that a construction that eliminates the location requirement from the previous

construction is the one most consistent with the Federal Circuit’s decision. The dispute on appeal

as to the location requirement was whether the centralized processing station is located at a cable-

TV network head-end unit (as construed by this Court and as Comcast argued) or whether it is

“location-agnostic” and instead “defined by its functionality of being ‘centralized’ with respect to

a multitude of users” (as Promptu argued). Corrected Principal Br. for Appellant 64, 66, Promptu

Sys. Corp. v. Comcast Corp., No. 22-1939 (Fed. Cir. Oct. 20, 2022). In urging the Federal Circuit

to reverse this Court’s construction, Prompu argued the term should be construed without reference

to its location, as “a portion of a cable television system at which voice commands are received

and command functions are transmitted.” Id. at 70.

In reversing on the location issue, the Federal Circuit rejected the position that the

centralized processing station must be located at a head-end unit. While the appellate court did

not explicitly adopt Promptu’s position, it also did not suggest that further exploration of the term

“centralized” should occur on remand, as it did when reversing other claim terms. See Promptu

Sys. Corp., 92 F.4th at 1379-80 (“back channel”); id. at 1381 (“speech recognition system”). The

guidance the court provided as to the meaning of “centralized” is largely consistent with Promptu’s

position. The appellate court characterized the term “centralized” as referring to the processing

station’s “relation to a group of set-top boxes,” rather than its location.21 Id. at 1383. And the

example the court provided of when centralization exists—although situating the processing

station “in the cable network”—also defines centralization in terms of the processing station’s

function of “serving multiple set-top boxes” rather than its location relative to them. See id. The

claim language reinforces this interpretation, as the claims themselves specify that the centralized

processing station receives and processes output from, and returns signals representing instructions

to, a “multitude” of set-top boxes. See ʼ538 patent col. 13 ll. 47-58.

Comcast argues the Federal Circuit’s decision makes clear that the centralized processing

station must be “in the cable network,” citing the appellate court’s statement that “[s]uch

centralization exists when the station is in the cable network, serving many set-top boxes, whether

or not it is located at the same place as a head-end unit.” See Defs.’ Br. 14, Dkt. No. 349. While

this is a plausible interpretation of the quoted sentence, notably, the appellate court did not suggest

that either the claim language or the specification requires the centralized processing station to be

located in the cable network. In this Court’s view, this statement is more plausibly understood as

an example of centralization than as a limitation on when centralization may exist.

Comcast also argues the intrinsic record supports the limitation that the centralized

procession station be “in the cable network,” but its arguments are not persuasive. Comcast asserts,

for example, that “[c]laim 34 recites that the centralized processing station is directly connected to

the ‘cable television link,’” which the parties agree is necessarily within the framework of a cable-

21 The Court recognizes that the Federal Circuit went on to cite the statement in Comcast’s

appellate brief that “[t]he claim term ‘centralized processing station’ makes clear that the

processing station is a discrete physical thing (a ‘station’) that is centrally located (‘centralized’)

vis-à-vis the ‘multitude of television set top boxes’ from which it receives signals.” Promptu Sys.

Corp., 92 F.4th at 1383. But the appellate court then provided an example of centralization that

did not require the processing station to be centrally located.

TV network. Defs.’ Responsive Br. 13-14, Dkt. No. 354. The claim recites that television set-top

boxes “transmit representative second output to a central processing station via the cable television

link” and that the centralized processing station “return[s] signals representing the instructions to

respective top boxes via the cable television link.” ʼ538 patent col. 13 ll. 44-58 (emphasis added).

While Comcast argues the cable television link is “the only physical medium on which the voice

commands travel from the set-top box to the processing station,” Defs.’ Responsive Br. 14, Dkt.

No. 354, the term “via” is not necessarily so restrictive. Notably, the only support Comcast offers

for this assertion is a description of a preferred embodiment in which voice commands travel from

a set-top box to a head-end unit “along a cable television uplink.” Id. (quoting ʼ538 patent col. 4

ll. 7-10). As the Federal Circuit recognized, however, this embodiment is merely exemplary. See

Promptu Sys. Corp., 92 F.4th at 1383.

Comcast also argues that Promptu’s current position that a centralized processing station

need not be in the cable network contradicts positions it took in the IPR proceedings. The evidence

Comcast offers does not support this assertion. Comcast first suggests Promptu agreed the

centralized processing station must be in the cable network by explaining the ʼ538 patent “was

directed at a ‘black box’ option meant to be plugged in to ‘existing cable industry network

architecture in the early 2000s.’” Defs.’ Responsive Br. 14, Dkt. No. 354 (quoting Defs.’ Ex. 11

at 10, Dkt. No. 354-2). The passage Comcast cites to support its position reveals that it was

Comcast’s expert (not Promptu) who asserted that implementing a voice recognition system for a

television network would simply require “buy[ing] the technology as a ‘black box’” or “buying an

‘off the shelf voice recognition system[].’” Defs.’ Ex. 11 at 10, Dkt. No. 354-2. Promptu

maintained the expert’s assertion was inaccurate and showed he lacked the experience necessary

to qualify as a person of ordinarily skill in the art, noting “[t]here were no ‘black box’ options

available” in the early 2000s. Id. Comcast also points to Promptu statement that the prior art’s

disclosure of “processing a voice command at a ‘remote server’” was “not enough to suggest a

centralized processing station, because a remote server is not necessarily either a ‘centralized

processing station’ or a ‘cable head-end server,’ even on a cable network.” Id. at 24. Comcast

argues Promptu’s proposed construction contradicts this statement, but, as Promptu notes, its

proposed construction does not equate a remote server with a centralized processing station, as a

centralized processing station must also satisfy the functional requirements described above.

The Court therefore construes “central/centralized processing station” to mean “a device

that receives voice commands, performs voice recognition on voice commands, and generates and

returns instructions to set-top boxes to carry out the commands.”

An appropriate order follows.

BY THE COURT:

/s/ Juan R. Sánchez

Juan R. Sánchez, J.