DETAILED ACTION
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/30/2026 has been entered.
Response to Amendment and Arguments.
Applicant’s arguments with respect to claim(s) 1-11, 13-16 and 18-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. The newly modified limitation “each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein one or more of the respective signs is associated with multiple matching functions; and” necessitates the new ground of rejection.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-7, 13-16, 18, 20 are rejected under 35 U.S.C. 103 as obvious over CN Patent CN 113656644 A, (ZHANG, HENG.) in view of US Patent US 20220327309 A1, (Carlock; John.) in view of US Patent US 20210174034 A1, (RETEK; David.) in further view of Cooper, Helen, and Richard Bowden. “Large lexicon detection of sign language.” International Workshop on Human-Computer Interaction. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007.
Claim 1
Regarding Claim 1, ZHANG. teach analyzing, during the communication session, the video to generate one or more feature vectors from the video data, each of the one or more feature vectors including values that represent a segment of a body of a person performing the sign language in the video data;
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 3 paragraph 5 from the bottom "extracting the image characteristic of the gesture training data set sample image by the backbone network in the teacher gesture detection model; determining the gesture label in the sample image through the classification prediction network in the teacher gesture detection model; and determining the gesture part in the sample image by the gesture frame prediction network in the teacher gesture detection model; "
Although feature vectors are not mentioned it is implicit since CNN's need vectors to operate. )
determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more feature vectors,
(Page 12 paragraph 12 "in step S31, obtaining the gesture training data set; each gesture data sample in the gesture training data set comprises a sample image, a gesture part label and a gesture label label."
Page 12 paragraph 13 and page 13 paragraph 1 wherein the gesture part label is the frame label of the gesture part in the sample image. the gesture data sample part of the gesture training data set is manually marked; the part is obtained by data enhancement processing based on the data sample of the manual marking; and the data sample based on the manual marking and the data sample obtained by the data enhancement processing are obtained by marking the unmarked sample image by means of semi-supervised mode. . The gesture tag refers to the gesture meaning represented by the gesture action."
Page 13 paragraph 5 "In one exemplary embodiment, according to the gesture training data set, the teacher gesture detection model for training, obtaining the training of the teacher gesture detection model, comprising: extracting the image characteristic of the gesture training data set sample image by the backbone network in the teacher gesture detection model; determining the gesture label in the sample image through the classification prediction network in the teacher gesture detection model; and determining the gesture part in the sample image by the gesture frame prediction network in the teacher gesture detection model; according to the gesture label, gesture part, gesture part label and gesture label label adjusting the network parameter of the teacher gesture detection model, obtaining the training of the teacher gesture detection model."
Page 9 paragraph 11 "The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 12 paragraph 2 "inputting a plurality of gesture label input language model obtained by the gesture detection model, coding the plurality of gesture labels through the encoder in the language model, obtaining the coding vector, and decoding the coding vector by the decoder in the language model, so as to translate the plurality of gesture tags into more smooth gesture language text information. wherein the gesture language represents the language represented by gesture action; it can include sign language or gesture dance and so on.")
determining, during the communication session, a second symbol from the first symbol for presentation during the communication session to facilitate communication.
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 2 paragraph 8-12 "
respectively gesture detection to the plurality of video frame pictures by the gesture detection model,
obtaining a plurality of gesture tags of the plurality of video frame pictures;
performing gesture language identification to the plurality of gesture tags through the language model to
obtain the gesture language text information;
outputting the gesture language text information.
Optionally, the output of the gesture language text information, comprising:
displaying the gesture language text information in the form of text;"
Page 12 Paragraph 12 "finally displaying the gesture language text information obtained by the translation to the user in the form of
text, or invoking the voice interface, converting the gesture language text information into voice information,
and playing out, finally achieving the purpose of communication.")
ZHANG. do not explicitly teach all of obtaining video data including sign language originating at a device during a communication session;
determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more feature vectors,
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein one or more of the respective signs is associated with multiple matching functions;and
a subset of the plurality of matching functions corresponding to sub-signs in the sign language, each of the sub-signs representing a gesture that forms part of a sign of the sign language;
However, Carlock teach obtaining video data including sign language originating at a device during a communication session;
(Paragraph 11 " A video communication system is disclosed comprising a plurality of video communication devices configured for hearing-impaired users to engage in communication sessions with hearing-capable users and a video relay service. The video relay service is configured to establish communication sessions between video communication devices associated with hearing-impaired users and far-end communication device associated with hearing-capable users, automatically generate translations of sign language content from a video stream from the corresponding video communication device during real-time communication sessions without a human sign language interpreter performing the sign language translation for the communication session, transmit the translation from the translation engine to the corresponding far-end communication device, automatically generate with the translation engine, a second translation of voice content from an audio stream from the corresponding far-end communication device, and transmit the second translation from the translation engine to the video communication device during the real-time communication session.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG to incorporate the teachings of Carlock to provide a “obtaining a first video data including sign language originating at a first device during a communication session;” Doing so would allow real time communication without the need of an interpreter, as recognized by Carlock. (Paragraph 8).
ZHANG in view of Carlock do not explicitly teach all of determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more feature vectors,
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein one or more of the respective signs is associated with multiple matching functions;and
a subset of the plurality of matching functions corresponding to sub-signs in the sign language, each of the sub-signs representing a gesture that forms part of a sign of the sign language;
However, RETEK teaches
determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more feature vectors,
(paragraph 111 " The task of the recognition layer 305 is to provide sign matches onto the phoneme/sign fragment stream, and as such into the time intervals of the sign utterance. Recognition and pattern matching of the signs can only be discussed together with the models and patterns trained behind. The features and matching algorithms encapsulating information that define how it is evaluated that a sign is matching a given interval of the utterance with some confidence are referred as sign patterns."
Paragraph 137 " In this way, a variety of different types of matching between signs and intervals of the aggregated segmentation may be provided through the graph of the phonemes/sign fragments. These may form a variety of possible sign sequences, all having confidence values either for the individual matched signs, and the transitions between the matched signs in the sequence. These possibilities may be regarded in a more abstract way as a directed acyclic graph (DAG) with the possible sign matches being nodes, and the possible succeeding signs (transitions) defining the (directed) edges. This graph is weighted both on its nodes with confidence of the match of the sign to the corresponding interval of features defined by the corresponding segments of the segmentation; and the edges with the confidence of the transition between the two signs and the corresponding features on the corresponding segments."
Feature and matching algorithm / sign patters is being interpreted as the matching function, confidence/confidence of the match is being interpreted as the similarity measure )
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein one or more of the respective signs is associated with multiple matching functions;and
(paragraph 111 " The task of the recognition layer 305 is to provide sign matches onto the phoneme/sign fragment stream, and as such into the time intervals of the sign utterance. Recognition and pattern matching of the signs can only be discussed together with the models and patterns trained behind. The features and matching algorithms encapsulating information that define how it is evaluated that a sign is matching a given interval of the utterance with some confidence are referred as sign patterns."
Paragraph 113 "Every sign pattern may include a weight distribution defining how important each feature channel is regarding that given sign. This means that some sign may be recognized based dominantly on the handshape and movement, while another based dominantly on handshape and location of the dominant hand relative to the face. This can be used accordingly in the evaluation and matching of the sign pattern to the sign fragments of given sign utterance. Signs and their patterns may be categorized into visually or algorithmically similar groups in the phase of training. This categorization may be done manually based on intentions incorporated in that sign, or algorithmically based on the distribution of features in the training data. This categorization provides information not only to the recognition part, but later to the grammatical parsing layer, where parsing can expect frequently confused signs, and may replace them with similar, but grammatically/semantically more fitting signs."
Paragraph 137 " In this way, a variety of different types of matching between signs and intervals of the aggregated segmentation may be provided through the graph of the phonemes/sign fragments. These may form a variety of possible sign sequences, all having confidence values either for the individual matched signs, and the transitions between the matched signs in the sequence. These possibilities may be regarded in a more abstract way as a directed acyclic graph (DAG) with the possible sign matches being nodes, and the possible succeeding signs (transitions) defining the (directed) edges. This graph is weighted both on its nodes with confidence of the match of the sign to the corresponding interval of features defined by the corresponding segments of the segmentation; and the edges with the confidence of the transition between the two signs and the corresponding features on the corresponding segments.")
wherein one or more of the respective signs is associated with multiple matching functions; and
(paragraph 115 "A given sign may be recognized based on a plurality of patterns. These alternative patterns may be defined algorithmically based on the training data. A typical example for a sign where it provide useful is the case of compound signs such as “TEACHER” which is a compound of “TEACH”+“PERSON”. In clear and formal signing this contains two separable movements for the two signs in the compound accordingly, but in natural signing these movements are merged into a single movement. Providing multiple patterns for this sign can incorporate both cases, where any decision of the segmentation can match to the appropriate pattern resulting a match in both cases.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock to incorporate the teachings of RETEK to provide a “determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more feature vectors, each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein one or more of the respective signs is associated with multiple matching functions; and” Doing so would Lead to a more accurate translation, as recognized by RETEK. (Paragraph 60).
ZHANG in view of Carlock in further view of RETEK do not explicitly teach all of
a subset of the plurality of matching functions corresponding to sub-signs in the sign language, each of the sub-signs representing a gesture that forms part of a sign of the sign language;
However, Cooper teach
a subset of the plurality of matching functions corresponding to sub-signs in the sign language, each of the sub-signs representing a gesture that forms part of a sign of the sign language;
(see table one shows smaller sign components.
page 1 abstract "…In the first, a set of viseme classifiers detects the presence of sub-Sign units of activity. The second level then assembles visemes into word level Sign using Markov chains. …"
Page 2 section 3 methodology " Sign language can be broken down into visemes in much the same way that speech can be broken down into phonemes. These visemes can be separated into 5 main categories [9] based on hand; shape(s) (dez), placement (tab), movement (sig), orientation(s) (ori) and arrangement (ha). This work concentrates on the tab, sig and ha visemes shown in table 1.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock in further view of RETEK to incorporate the teachings of Cooper to provide a “a subset of the plurality of matching functions corresponding to sub-signs in the sign language, each of the sub-signs representing a gesture that forms part of a sign of the sign language;” Doing so would Make the system able to cope a large lexicon and more expandable, as recognized by Cooper. (abstract).
Claim 2
Regarding Claim 2, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Zhang teaches the method of claim 1, wherein the first
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 12 paragraph 2 "inputting a plurality of gesture label input language model obtained by the gesture detection model, coding the plurality of gesture labels through the encoder in the language model, obtaining the coding vector, and decoding the coding vector by the decoder in the language model, so as to translate the plurality of gesture tags into more smooth gesture language text information. wherein the gesture language represents the language represented by gesture action; it can include sign language or gesture dance and so on.")
Claim 3
Regarding Claim 3, ZHANG in view of Carlock in view of RETEK, in further view of Cooper, furthermore Zhang teaches
The method of claim 1, wherein the second
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language.")
Claim 4
Regarding Claim 4, ZHANG in view of Carlock in view of RETEK, in further view of Cooper, furthermore Zhang teaches
The method of claim 1, wherein the language model uses gloss.
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language.")
Claim 5
Regarding Claim 5, ZHANG in view of Carlock in view of RETEK, in further view of Cooper, furthermore Zhang teaches the method of claim 1, wherein determining the[[a]] second
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language.")
Claim 6
Regarding Claim 6, ZHANG in view of Carlock in view of RETEK, in further view of Cooper, furthermore Zhang teaches the method of claim 1, further comprising providing the second
(Page 2 paragraph 11-13 "Optionally, the output of the gesture language text information, comprising:
displaying the gesture language text information in the form of text; or
converting the gesture language text information into voice information, and playing the voice information.")
Claim 7
Regarding Claim 7, ZHANG in view of Carlock in view of RETEK, in further view of Cooper, furthermore Zhang teaches
The method of claim [[1]]6, wherein the second symbol is provided as audio
(Page 2 paragraph 11-13 "Optionally, the output of the gesture language text information, comprising:
displaying the gesture language text information in the form of text; or
converting the gesture language text information into voice information, and playing the voice information."
Page 12 Paragraph 12 "finally displaying the gesture language text information obtained by the translation to the user in the form of
text, or invoking the voice interface, converting the gesture language text information into voice information,
and playing out, finally achieving the purpose of communication.")
Claim 13
Regarding Claim 13, ZHANG teach analyzing, during the communication session, the video to generate one or more feature vectors from the video data, each of the one or more feature vectors including values that represent a segment of a body of a person performing the sign language in the video data;
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 3 paragraph 5 from the bottom "extracting the image characteristic of the gesture training data set sample image by the backbone network in the teacher gesture detection model; determining the gesture label in the sample image through the classification prediction network in the teacher gesture detection model; and determining the gesture part in the sample image by the gesture frame prediction network in the teacher gesture detection model; "
Although feature vectors are not mentioned it is implicit since CNN's need vectors to operate. )
determining, during the communication session, a first symbol from the one or more matching functions to facilitate communication during the communication session.
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 2 paragraph 8-12 "respectively gesture detection to the plurality of video frame pictures by the gesture detection model, obtaining a plurality of gesture tags of the plurality of video frame pictures; performing gesture language identification to the plurality of gesture tags through the language model to obtain the gesture language text information; outputting the gesture language text information. Optionally, the output of the gesture language text information, comprising: displaying the gesture language text information in the form of text;"
Page 12 Paragraph 12 "finally displaying the gesture language text information obtained by the translation to the user in the form of text, or invoking the voice interface, converting the gesture language text information into voice information, and playing out, finally achieving the purpose of communication.")
ZHANG. do not explicitly teach all of
obtaining video data including sign language originating at a device during a communication session;
determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more features vectors,
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function,
wherein at least one of the plurality of matching functions corresponds to a part of a first sign and a part of a second sign and one or more of the respective signs are associated with multiple matching functions; and
However, Carlock teach obtaining video data including sign language originating at a device during a communication session;
(Paragraph 11 " A video communication system is disclosed comprising a plurality of video communication devices configured for hearing-impaired users to engage in communication sessions with hearing-capable users and a video relay service. The video relay service is configured to establish communication sessions between video communication devices associated with hearing-impaired users and far-end communication device associated with hearing-capable users, automatically generate translations of sign language content from a video stream from the corresponding video communication device during real-time communication sessions without a human sign language interpreter performing the sign language translation for the communication session, transmit the translation from the translation engine to the corresponding far-end communication device, automatically generate with the translation engine, a second translation of voice content from an audio stream from the corresponding far-end communication device, and transmit the second translation from the translation engine to the video communication device during the real-time communication session.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG to incorporate the teachings of Carlock to provide a “obtaining a first video data including sign language originating at a first device during a communication session;” Doing so would allow real time communication without the need of an interpreter, as recognized by Carlock. (Paragraph 8).
ZHANG in view of Carlock do not explicitly teach all determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more features vectors,
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function,
wherein at least one of the plurality of matching functions corresponds to a part of a first sign and a part of a second sign and one or more of the respective signs are associated with multiple matching functions; and
However, RETEK teaches
determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more features vectors,
(paragraph 111 " The task of the recognition layer 305 is to provide sign matches onto the phoneme/sign fragment stream, and as such into the time intervals of the sign utterance. Recognition and pattern matching of the signs can only be discussed together with the models and patterns trained behind. The features and matching algorithms encapsulating information that define how it is evaluated that a sign is matching a given interval of the utterance with some confidence are referred as sign patterns."
Paragraph 137 " In this way, a variety of different types of matching between signs and intervals of the aggregated segmentation may be provided through the graph of the phonemes/sign fragments. These may form a variety of possible sign sequences, all having confidence values either for the individual matched signs, and the transitions between the matched signs in the sequence. These possibilities may be regarded in a more abstract way as a directed acyclic graph (DAG) with the possible sign matches being nodes, and the possible succeeding signs (transitions) defining the (directed) edges. This graph is weighted both on its nodes with confidence of the match of the sign to the corresponding interval of features defined by the corresponding segments of the segmentation; and the edges with the confidence of the transition between the two signs and the corresponding features on the corresponding segments."
Feature and matching algorithm / sign patters is being interpreted as the matching function, confidence/confidence of the match is being interpreted as the similarity measure )
each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function,
(paragraph 111 " The task of the recognition layer 305 is to provide sign matches onto the phoneme/sign fragment stream, and as such into the time intervals of the sign utterance. Recognition and pattern matching of the signs can only be discussed together with the models and patterns trained behind. The features and matching algorithms encapsulating information that define how it is evaluated that a sign is matching a given interval of the utterance with some confidence are referred as sign patterns."
Paragraph 113 "Every sign pattern may include a weight distribution defining how important each feature channel is regarding that given sign. This means that some sign may be recognized based dominantly on the handshape and movement, while another based dominantly on handshape and location of the dominant hand relative to the face. This can be used accordingly in the evaluation and matching of the sign pattern to the sign fragments of given sign utterance. Signs and their patterns may be categorized into visually or algorithmically similar groups in the phase of training. This categorization may be done manually based on intentions incorporated in that sign, or algorithmically based on the distribution of features in the training data. This categorization provides information not only to the recognition part, but later to the grammatical parsing layer, where parsing can expect frequently confused signs, and may replace them with similar, but grammatically/semantically more fitting signs."
Paragraph 137 " In this way, a variety of different types of matching between signs and intervals of the aggregated segmentation may be provided through the graph of the phonemes/sign fragments. These may form a variety of possible sign sequences, all having confidence values either for the individual matched signs, and the transitions between the matched signs in the sequence. These possibilities may be regarded in a more abstract way as a directed acyclic graph (DAG) with the possible sign matches being nodes, and the possible succeeding signs (transitions) defining the (directed) edges. This graph is weighted both on its nodes with confidence of the match of the sign to the corresponding interval of features defined by the corresponding segments of the segmentation; and the edges with the confidence of the transition between the two signs and the corresponding features on the corresponding segments.")
wherein at least one of the plurality of matching functions corresponds to a part of a first sign and a part of a second sign and one or more of the respective signs are associated with multiple matching functions; and
(paragraph 73 " Ultimately, the unit of translation is not a single gesture/sign, but a whole utterance of signing. This is may be a sentence, but does not necessarily correspond to the grammatical definition of a sentence. Similar to a situation where an interpreter interprets for the signer, the signer produces a whole utterance, before the interpreter translates it. This is important, as even the signs may have different meaning based on the succeeding signs in the same sentence. Another aspect is that the translated language (e.g., English in one embodiment) may have different word order, so latter signs in the signed sentence may completely rearrange the word order in the translated sentence. A third (and maybe most interesting) case is when a sign is technically the same as a part of a different sign. For example, the first part of the sign “GRANDFATHER” is the sign “FATHER”, and the second half of it is extremely similar to the sign “WILL”, so any occurrence of . . . GRANDFATHER . . . will look very similar to . . . FATHER WILL . . . In a similar fashion AGREENOT (having the English meanings both “do-not-agree” and “disagreement”) is very similar to ANSWER+BUT. In this case only the other parts of the sentence can resolve the ambiguity of the sign recognition with the help of grammatical parsing."
Paragraph 135 "An aggregate segmentation 410 is extracted by the feature extraction layer 304, with confidences of the boundaries. There may be aggregations across these features, e.g. the features of the non-dominant hand may be aggregated based on the segmentation of the dominant hand, as this often provides relevant result the sign being structured by the dominant hand. Possible suggested concatenations of the segments are also provided based on the confidences and other features of the aggregated segments. This means that segments are not a strictly linear structure. A big arc of hand movement may be treated as a single longer segment (segment of 425D), but may also be treated as concatenation of two smaller arcs (segments of 425B and 425C). Features aggregated at the intervals and the boundaries of the possibly aggregated segments together with these segments form the phonemes/sign fragments. In the example the bigger arc and the smaller arcs all provide their respective phonemes/sign fragments (425B, 425C, 425D). This also means that the intervals of these phonemes/sign fragments may intersect. These phonemes/sign fragments 425 may be extracted based on the appropriate aggregation of features (404) on a plurality of intervals of the aggregated segmentation (410). These phonemes/sign fractures naturally form a graph with edges pointing to the possible succeeding phonemes/sign fragments (425A to 425B and 425D; from 425B to 425C, etc. as indicated by the arrows). There may be skipped intervals (e.g., between 425E and 425G) of the aggregated segments, and some phonemes/sign fragments may not have successors besides the phonemes/sign fragments at the end of the utterance (e.g., 425G). These phonemes/sign fragments are used in the matching of signs onto the intervals of the feature stream by the recognition layer 305. Signs detected are matched to parts of paths in the graph of phonemes/sign fragments (e.g. sign GRANDFATHER (418) is matched to the path 425A-425B-425C). Sign matchings 305 are evaluated based on these sign fragments. A sign may match to one or more concatenated sign fragments. Some sign fragments may be omitted because they are treated either as transitional, or as noise 422. Special sign roles may also be decided (e.g., pause, inactivity, etc.). These do not correspond to real signs, but intervals of sign fragments matching these intervals of the sign utterance (e.g., pause, sections 420 between sentences, noise, etc.) containing significant information."
paragraph 115 "A given sign may be recognized based on a plurality of patterns. These alternative patterns may be defined algorithmically based on the training data. A typical example for a sign where it provide useful is the case of compound signs such as “TEACHER” which is a compound of “TEACH”+“PERSON”. In clear and formal signing this contains two separable movements for the two signs in the compound accordingly, but in natural signing these movements are merged into a single movement. Providing multiple patterns for this sign can incorporate both cases, where any decision of the segmentation can match to the appropriate pattern resulting a match in both cases.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock to incorporate the teachings of RETEK to provide a “determining, during the communication session, a similarity measure for each of a plurality of matching functions using the one or more features vectors,each of the plurality of matching functions associated with at least one a respective sign of the sign language and each of the plurality of matching functions configured to compute a similarity measure indicating a correspondence between feature vectors and a sign associated with the corresponding matching function, wherein at least one of the plurality of matching functions corresponds to a part of a first sign and a part of a second sign and one or more of the respective signs are associated with multiple matching functions; and” Doing so would Lead to a more accurate translation, as recognized by RETEK. (Paragraph 60).
Claim 14
Regarding Claim 14, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Zhang teaches the method of claim 13, further comprising translating the first
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language.")
Claim 15
Regarding Claim 15, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Zhang teaches the method of claim 13, wherein the first
(page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language."
Page 12 paragraph 2 "inputting a plurality of gesture label input language model obtained by the gesture detection model, coding the plurality of gesture labels through the encoder in the language model, obtaining the coding vector, and decoding the coding vector by the decoder in the language model, so as to translate the plurality of gesture tags into more smooth gesture language text information. wherein the gesture language represents the language represented by gesture action; it can include sign language or gesture dance and so on." Gloss
page 9 paragraph 11"The embodiment of the present disclosure performs frame extraction processing to the video to be identified, obtaining a plurality of extracted video frame pictures; performing gesture detection on a plurality of video frame pictures by a gesture detection model, respectively a plurality of gesture labels of a plurality of video frame pictures; performing gesture language recognition for a plurality of gesture labels through the language model; obtaining the gesture language text information; outputting gesture language text information; because the gesture detection model and the language model are matched to identify the gesture language of the video, realizing the real-time identification of the end to the end; it can improve the identification efficiency of the gesture language." Script)
Claim 16
Regarding Claim 16, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Zhang teaches 16. The method of claim 14, further comprising:
generating a
(Page 2 paragraph 11-13 "Optionally, the output of the gesture language text information, comprising: displaying the gesture language text information in the form of text; or converting the gesture language text information into voice information, and playing the voice information.")
providing the
(Page 2 paragraph 11-13 "Optionally, the output of the gesture language text information, comprising:
displaying the gesture language text information in the form of text; or
converting the gesture language text information into voice information, and playing the voice information.")
Claim 18
Regarding Claim 18, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Carlock teaches the method of claim 13 [[17]], wherein the gesture a _third sign includes a third
(Paragraph 57-59 "In embodiments where the translation if from sign language content to word content in either a single phone or multiple phone system, the translation engine is configured to identify at least one sign language content segment in a video stream. A sign language content segment may include chunks of data or video frames that contain the arm and/or hand motions for a particular letter, word, or phrase in sign language. The sign language content segments may be determined by observing or identifying pauses that typically occur between arm and/or hand motions when a signer signs a particular letter, word, or phrase in sign language. The pauses may be small in some cases. In one embodiment, the translation engine may be trained to identify pauses by observing signers signing over an extended period of time.
The translation engine may also be configured to identify at least one content indicator corresponding to the sign language content segment. In one embodiment, translation of sign language content may depend upon recognizing more than just the hand and/or arm motion comprising the sign language content. Content indicators include visual cues provided by the motion, expression, and body language of the signer that are in addition to the minimum hand and arm movement required to convey particular sign language content. These content indicators provide a better translation of sign language content into word content. Certain content indicators also provide punctuation for a language that otherwise has none.
When a signer performs the arm and hand motions representing sign language content, they may do it with a smile, or scowl, or speed of motion, or other visual que that helps provide meaning to the sign language content. In one embodiment, a content indicator comprises one or more of a head position, a head motion, a face position, a face motion, a body position, a body motion, an arm motion speed, a hand motion speed, an arm motion range, and a hand motion range. A content indicator may include one or more of a lip position or motion, a cheek position, an eyebrow position, an eye opening size, a brow position or motion, a mouth position or motion and the like. Content indicators such as facial expressions and body language convey an emotion, feeling, or mood of the signer such as, by way of nonlimiting example, love, happiness, sadness, fear, anger, disgust, surprise, acceptance, anticipation, shock, disappointment, disbelief, energy, melancholy, and eagerness to name just a few. These emotions, feelings, or moods provide context beyond the lingual context of a particular sentence, that help provide a more accurate translation for the sign language content provided by the signer displaying such content indicators while signing the sign language content."
Paragraph 63 "The content indicators also allow for the expression of varying degrees of something using the same sign. For example, the sing for small is indicated by bringing the palms of both hands toward each other. However, if the shoulders are rolled forward and the area or space typically used to sign the word small is decreased, the signee may use the motion representing “small,” but may actually mean “very small” or “even smaller.” Additionally, the sign language content representing “rain” is made by using the hands slightly above the shoulders, palms facing out, fingers extended and pivoting the hands downward at the wrist. However, if this sign language content is expressed more animatedly, with a certain abruptness to the motion, the user or signer is more than likely intending to express a lot of rain or a downpour. A smaller, slower motion for this sign probably means less than your typical rain and more of a sprinkle."
examiner notes the first sign is the sign representing rain, second sign is the sign representing a lot of rain (down pour) the third sign represents less than rain (a sprinkle))
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock in view of RETEK, in further view of Cooper to incorporate the teachings of Carlock to provide a “The method of claim 17, wherein the second part of the first sign includes a first one or more states, the first part of the second sign includes a second one or more states, and the third sign includes a third one or more states.” Doing so would not only examine arm and hand motion but as well as visual cues to efficiently translate., as recognized by Carlock. (Paragraph 1).
Claim 20
Regarding Claim 20, ZHANG in view of Carlock in view of RETEK, in further view of Cooper. furthermore, Carlock teaches the method of claim 18, wherein: second gesture gesture [[sign]] are the same gesture[[state]].
(Paragraph 63 "The content indicators also allow for the expression of varying degrees of something using the same sign. For example, the sing for small is indicated by bringing the palms of both hands toward each other. However, if the shoulders are rolled forward and the area or space typically used to sign the word small is decreased, the signee may use the motion representing “small,” but may actually mean “very small” or “even smaller.” Additionally, the sign language content representing “rain” is made by using the hands slightly above the shoulders, palms facing out, fingers extended and pivoting the hands downward at the wrist. However, if this sign language content is expressed more animatedly, with a certain abruptness to the motion, the user or signer is more than likely intending to express a lot of rain or a downpour. A smaller, slower motion for this sign probably means less than your typical rain and more of a sprinkle.")
See claim 18 for rationale.
Claims 8 are rejected under 35 U.S.C. 103 as obvious over CN Patent CN 113656644 A, (ZHANG, HENG.) in view of US Patent US 20220327309 A1, (Carlock; John.) in view of US Patent US 20210174034 A1, (RETEK; Nikolas Anthony.) in view of Cooper, Helen, and Richard Bowden. “Large lexicon detection of sign language.” International Workshop on Human-Computer Interaction. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007. in further view of CN patent CN 114758411 A , (LI, Tian-jun )
Claim 8
Regarding Claim 8, ZHANG in view of Carlock in view of RETEK, in further view of Cooper do not explicitly teach all of the method of claim 1, wherein the language model includes a statistical language model.
However, Tian-jun teach the method of claim 1, wherein the language model includes a statistical language model.
(Page 4 paragraph 4 "According to one aspect of the present invention, the output of the model is decoded into a natural language sequence, comprising: decoding the probability vector output by the decoder into the natural language sequence, the combined natural language model and the probability of the sign language-natural language translation model, predicting the local optimal natural language sequence by the beam search method. wherein the probability of the sign language-natural language translation model is the probability vector sequence output by the decoder, the natural language model is a natural language model based on n-gram (n-gram model), the probability vector for decoding the word sequence will be the weighted average of the probability vector output by the translation model and the probability corresponding to the language model; to obtain balance on the optimal solution and decoding efficiency, using beam search method obtain local optimal natural language word sequence.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock in view of RETEK, in further view of Cooper to incorporate the teachings of Tian-jun to provide a “The method of claim 1, wherein the language model includes a statistical language model.” Doing so would make the model find the best corresponding path , as recognized by Tian-jun . (page 2 paragraph 3).
Claims 9 are rejected under 35 U.S.C. 103 as obvious over CN Patent CN 113656644 A, (ZHANG, HENG.) in view of US Patent US 20220327309 A1, (Carlock; John.) in view of US Patent US 20210174034 A1, (RETEK; Nikolas Anthony.) in view of Cooper, Helen, and Richard Bowden. “Large lexicon detection of sign language.” International Workshop on Human-Computer Interaction. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007. in further view of CN patent CN 109032356 B, (QU, Xiao-feng)
Claim 9
Regarding Claim 9, ZHANG in view of Carlock in view of RETEK, in further view of Cooper do not explicitly teach all of the method of claim 1, wherein the language model uses at least one neural network.
However, Xiao-feng teach the method of claim 1, wherein the language model uses at least one neural network.
(Page 9-10 paragraph 14 and 15 "step 470, when the neural network model convergence, the convergence neural network the model as the sign language action recognition by the sign language action of the recognition language action recognition the image data.
if the iteration times reaches the maximum, or mathematical mapping relation reaches the optimal, then neural network the convergence of the model, namely the convergence neural network the model as a recognition language model, executing image data sign language action recognition")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock in view of RETEK, in further view of Cooper to incorporate the teachings of Xiao-feng to provide a “The method of claim 1, wherein the language model uses at least one neural network.” Doing so would make it have more long-term memory, as recognized by Xiao-feng. (page 8 paragraph 15).
Claims 10 and 11 are rejected under 35 U.S.C. 103 as obvious over CN Patent CN 113656644 A, (ZHANG, HENG.) in view of US Patent US 20220327309 A1, (Carlock; John.) in view of US Patent US 20210174034 A1, (RETEK; Nikolas Anthony.) in view of Cooper, Helen, and Richard Bowden. “Large lexicon detection of sign language.” International Workshop on Human-Computer Interaction. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007. in further view of US patent US 20230409842 A1, (Ati; Modafar Kadim)
Claim 10
Regarding Claim 10, ZHANG in view of Carlock in view of RETEK, in further view of Cooper do not explicitly teach all of the method of claim 1, further comprising determining a third
However, Modafar Kadim teach the method of claim 1, further comprising determining a third set of one or more symbols from the second set of one or more symbols.
(paragraph 35 "At step 402, text language associated with spoken language A is inputted into electronic application 203 via a graphical user interface that is displayed on a user device (e.g., user device 202, user device 204). At step 404, electronic application 203 analyze the inputted text language and may split the inputted text language into separate words so that each word can be translated individually. At step 406, electronic application 203 and/or translation system 206 (which is a part of electronic application 203 or is in communication with electronic application 203) may translate the text language associated with spoken language A into the text language associated with spoken language B.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified ZHANG in view of Carlock in view of RETEK, in further view of Cooper to incorporate the teachings of Modafar Kadim to provide a “The method of claim 1, further comprising determining a third set of one or more symbols from the second set of one or more symbols.” Doing so would make translation occur for individuals remotely without the need for a translator present, as recognized by Modafar Kadim. (Paragraph 16).
Claim 11
Regarding Claim 11, ZHANG. In view of Carlock in further view of Modafar Kadim, furthermore Modafar Kadim teaches the method of claim 10, wherein determining a third
(paragraph 35 "At step 402, text language associated with spoken language A is inputted into electronic application 203 via a graphical user interface that is displayed on a user device (e.g., user device 202, user device 204). At step 404, electronic application 203 analyze the inputted text language and may split the inputted text language into separate words so that each word can be translated individually. At step 406, electronic application 203 and/or translation system 206 (which is a part of electronic application 203 or is in communication with electronic application 203) may translate the text language associated with spoken language A into the text language associated with spoken language B.")
See claim 10 for rationale.
Claims 19 are rejected under 35 U.S.C. 103 as obvious over CN Patent CN 113656644 A, (ZHANG, HENG.) in view of US Patent US 20220327309 A1, (Carlock; John.) in view of US Patent US 20210174034 A1, (RETEK; David.) in view of Cooper, Helen, and Richard Bowden. “Large lexicon detection of sign language.” International Workshop on Human-Computer Interaction. Berlin, Heidelberg: Springer Berlin Heidelberg, 2007 in further view of US 20160307469 A1 (Zhou; Zhengyu)
Claim 19
Regarding Claim 19, ZHANG in view of Carlock in view of RETEK, in further view of Cooper do not explicitly teach all of The method of claim 18, wherein thesecond gesture is tied to the third gesture such that training of the second gesture results in training of the third gesture.
However, Zhou teaches The method of claim 18, wherein thesecond gesture is tied to the third gesture such that training of the second gesture results in training of the third gesture.
(Paragraph 27 "As used herein, the terms “sign phoneme” or more simply “phoneme” are used interchangeably and refer to a gesture, handshape, location, palm orientation, or other hand posture using one or two hands that corresponds to the smallest unit of a sign language. Each sign in a predetermined sign language includes at least one sign phoneme, and as used herein the term “sign” refers to a word or other unit of language that is formed from one or more sign phonemes. In some instances a single sign phoneme corresponds to a single word, while in other instances multiple phonemes form a single word depending upon the conventions of the sign language. Many signed words can use the same phoneme or different phonemes more than once in a sequence to form a single word or express an idea in the context of a larger sentence or phrase. As used herein, the term “transition” refers to the movements of the hands that are made between individual signs, such as between sequences of sign phonemes that form individual words or between the phonemes within a single sign. Unlike a sign phoneme, a transition movement is not a typical unit of a sign language that has independent meaning, but transition movements still provide important contextual information to indicate the end of a first word or phoneme and the beginning of a second word or phoneme. As used herein, the term “st/end” refers to hand movements and postures of the user for starting and ending a sequence of signs, including predetermined sequences of signs that are used in a training process for sign language recognition and for non-predetermined sequences of signs that an automated sign language recognition system detects during operation. As used herein, the term “phrase” refers to any sequence of signs that include some form of start and end hand motions along with motions for at least one individual sign. In phrases that include multiple signs, a transition hand movement separates individual signs within the phrase. Phrases can correspond to any sequence of interrelated signs such as signs corresponding to multiple words in a complete sentence or shorter word sequence. A phrase can also include, for example, sequences of numbers or individual letters in a predetermined sign language. "
Paragraph 46 " For sign languages, a word may contain one or more sign phonemes. The system 100 generates the HMM 120 for the sign phonemes to accurately model words based on three requirements. First, words have to be modeled in order to evaluate the feature probability, P(X|W). Second, language models have to be used to calculate the prior sentence probability, P(W). Third, the word models and language models have to be combined to form networks of HMM states in which the decoding for the most likely word string W* in the search space Γ can be performed. During the process 800, the processor 104 performs the training process to generate the HMM 120 for the sign phonemes based on the observed features for the signs conditioned upon the signs in the predetermined training sequences of signs. The processor 104 similarly performs the training process for the transitions 124 based on the observed features for the transitions conditioned upon the transitions between signs in the predetermined training sequences of signs and the training process for the phrase st/end hand movements 116 based on the observed features for the st/end hand movements conditioned upon the st/end hand movements in the predetermined training data 130.")
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified , ZHANG in view of Carlock in view of RETEK, in further view of Zhou to incorporate the teachings of Cooper to provide a “The method of claim 18, wherein the second gesture is tied to the third gesture such that training of the second gesture results in training of the third gesture.” Doing so would Detect non-predetermined sequences of signs, as recognized by Zhou. (Paragraph 27).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALI M HASSAN whose telephone number is (571)272-5331. The examiner can normally be reached Monday - Friday 8:00am - 4:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Paras Shah can be reached at (571)270-1650. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ALI M HASSAN/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
06/09/2026
Prosecution Insights