DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This Office Action is in response to correspondence field 21 July 2026 in reference to application 18/858,696. Claims 1-17 are pending and have been examined.
Response to Amendment
The amendment filed 21 July 2026 has been accepted and considered in this office action. Claims 1-15 have been amended and claims 16 and 17 added.
Claim Objections
Claim 11 is objected to because of the following informalities: An extra “w” appears at the beginning of line 2. Appropriate correction is required.
Response to Arguments
Applicant’s arguments, see Remarks, filed 21 July 2026, with respect to rejections made under 35 USC 101 have been fully considered and are persuasive. The 35 USC 101 rejection of the claims has been withdrawn.
Applicant’s arguments with respect to claim(s) 1-15 and the prior art 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.
Claim Rejections - 35 USC § 103
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1, 11, 15, and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sohn et al. (US PAP 2020/0051572) in view of Khoury et al. (US PAP 2021/0326421).
Consider claim 1, Sohn teaches An artificial intelligence apparatus (abstract, figure 1) comprises:
a memory storing a list of pre-learned speakers (0058, memory storing account information for registered users, 0047-49, registered users voiceprints and account information); and,
a processor (0057) configured to:
preprocess utterance data when the utterance data is input to generate preprocessed utterance data (0053, noise filtering, feature extraction, etc.);
identify a new speaker based on the pre-processed utterance data (0033, 0062-63, determining that voiceprint is associated with a user who is not registered),
output an active query to the new speaker (0068, system queries user if they would like to register), and
when response utterance data of the new speaker to the active query is input, learn the new speaker based on the response utterance data and register the new speaker in the list of pre-learned speakers (0069-74, user may register by conversational inputs, and account created as a registered user.).
Sohn does not specifically teach:
inputting the preprocessed utterance data into a neural network model,
calculating an uncertainty score for the utterance data using the neural network model,
identifying a speaker of the utterance data as a new speaker if the uncertainty score is greater than or equal to a reference value.
In the same field of speaker identification, Khoury teaches
inputting the preprocessed utterance data into a neural network model (0112, identification of speakers may be performed via a neural network model),
calculating an uncertainty score for the utterance data using the neural network model (0142-44,similarity scores for matches, mathematically the inverse of uncertainty, but functionally equivalent ),
identifying a speaker of the utterance data as a new speaker if the uncertainty score is greater than or equal to a reference value (0145, if similarity falls below threshold, i.e. greater uncertainty, new speaker profile is created ).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to use a determine that a new speaker is present based on similarity (and thus uncertainty) scores as taught by Khoury in the system of Sohn in order to allow the system to accurately determine if a new speaker is present.
Consider claim 11, Sohn teaches Claim the artificial intelligence apparatus of claim 1,
wherein the processor, when outputting the active query, selects a specific active query corresponding to the new speaker from a list of pre-stored active queries based on the uncertainty score, and outputs the specific active query to the new speaker (0036-37, queries to user may be selected from pre-determined actions, 0068 different queries for registration. In combination with would only occur if the similarity score was low).
Consider claim 15, Sohn teaches A method for automatically recognizing speaker of an artificial intelligence apparatus (abstract), comprising:
receiving, by a processor of the artificial intelligence apparatus (0057), utterance data from a speaker; (0060, voice signal uttered by user);
preprocessing, by the processor, utterance data to generate preprocessed utterance data (0053, noise filtering, feature extraction, etc.) (0053, noise filtering, feature extraction, etc.),
identifying a new speaker based on the pre-processed utterance data (0033, 0062-63, determining that voiceprint is associated with a user who is not registered),
outputting, by the processor, an active query to the new speaker (0068, system queries user if they would like to register);
receiving, by the processor, response utterance data from the new speaker to the active query (0068-69, receiving responses from a user);
learning, by the processor, the new speaker based on the response utterance data of the new speaker (0069-74, user may register by conversational inputs, and account created as a registered user.) and
registering, by the processor, the new speaker in the speaker list (0069-74, user may register by conversational inputs, and account created as a registered user).
Sohn does not specifically teach:
inputting the preprocessed utterance data into a neural network model,
calculating an uncertainty score for the utterance data using the neural network model,
identifying a speaker of the utterance data as a new speaker if the uncertainty score is greater than or equal to a reference value.
In the same field of speaker identification, Khoury teaches
inputting the preprocessed utterance data into a neural network model (0112, identification of speakers may be performed via a neural network model),
calculating an uncertainty score for the utterance data using the neural network model (0142-44,similarity scores for matches, mathematically the inverse of uncertainty, but functionally equivalent ),
identifying a speaker of the utterance data as a new speaker if the uncertainty score is greater than or equal to a reference value (0145, if similarity falls below threshold, i.e. greater uncertainty, new speaker profile is created ).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to use a determine that a new speaker is present based on similarity (and thus uncertainty) scores as taught by Khoury in the system of Sohn in order to allow the system to accurately determine if a new speaker is present.
Consider claim 17, Sohn and Khoury teaches the artificial intelligence apparatus of claim 11, wherein the processor increases a number of active queries selected from the list of pre- stored active queries as the uncertainty score decreases (Sohn 0067, when new speaker detected, i.e. low similarity in Khoury, responses are limited to registration, not the other responses contemplated in 0036).
Claim(s) 2-7, 12, and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sohn and Khoury as applied to claim 1 above, and further in view of Kim et al. (Message Passing Adaptive Resonance Theory for Active Semi-Supervised learning).
Consider claim 2, Sohn and Khoury teach the artificial intelligence apparatus of claim 1, wherein the processor, when preprocessing the utterance data, performs feature extraction from the utterance data (0053, feature extraction)
Sohn and Khoury do not specifically teach performing dimension reduction.
In the same field of data classification, Kim teaches performing dimension reduction (figure 1, section 5.3, dimension reduction).
It would have been obvious to one or ordinary skill in the art at the time of effective filing to reduce dimensions as taught by Kim in the system of Sohn and Khoury in order to reduce computational costs for processing the inputs.
Consider claim 3, Sohn and Khoury and Kim teach The artificial intelligence apparatus of claim 2. Kim further teaches wherein the processor, when identifying the speaker of the utterance data, inputs the preprocessed utterance data into the neural network model to configure a first node corresponding to the utterance data in an embedding space (section 4.1, node formation, corresponding to input data embeddings. In combination with Sohn the input data would be speech, “Node classification” specifically creates new nodes), and connects an edge between the first node and a second node that already exists in the embedding space based on a correlation between nodes (section 4.1, edge formation, connecting nodes based on correlation), and calculates the uncertainty score based on the connection relationship of the edge (section 4.1, initiating a new node if no node activates for a new label. In Sohn this would be a new speaker).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to apply the MPART learning methods for classifying data as taught by Kim in the speaker classification system of Sohn in order to more accurately learn unclassified data on the fly (Kim Introduction).
Consider claim 4, Kim teaches The artificial intelligence apparatus of claim 3,
wherein the processor, when configuring the first node, if the utterance data does not satisfy a similarity criterion condition with a data group of the second node where utterance data of an existing speaker already exists, configures the first node to include the utterance data (section 4.1, where no classification nodes activated based on similarity, new node is created. In combination with Sohn, the labels would correspond to speakers), and if the c utterance data satisfies the similarity criterion condition with the data group of the second node, includes the utterance data in the data group of the second node (section 4.1, node creation, matching similarity and selecting a winning node if the noes are activated).
Consider claim 5, Kim teaches The artificial intelligence apparatus of claim 3, wherein the processor, when connecting the first node and the second node with the edge, calculates a weight based on a co-activated count between the first node and the second node, and connects the first node and the second node with the edge based on the weight (section 4.1, edge formation, weights calculated based on co-activations show in equation 3).
Consider claim 6, Kim teaches The artificial intelligence apparatus of claim 5,
wherein the processor does not connect the first node and the second node with the edge if the weight is 0 (Section 4.1, edge formation, if weight is zero, the nodes are not co-activated and not connected).
Consider claim 7, Kim teaches The artificial intelligence apparatus of claim 5,
wherein the processor, if a similarity between the first node and the second node is high, increases the co-activated count between the first node and the second node is increased (section 4.1, node formation, as similarity increased, activation increase.).
Consider claim 12, Sohn teaches the artificial intelligence apparatus of claim 1,
wherein the processor, when the response utterance data of the new speaker to the active query is input, checks whether the response utterance data satisfies the active query, and if the response utterance data satisfies the active query, labels the response data to learn the new speaker, and registers the new speaker in the list of pre-learned speakers (0074, when information required to register speaker is in present from conversational replies, system may register voiceprint from utterances and create new user registration.).
Sohn does not specifically teach labelling the utterance data included in the first node to learn the new speaker.
In the same field of data classification, Kim teaches labelling the utterance data included in the first node to learn the new speaker (section 4.2, node classification, labels are received for each created node).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to apply the MPART learning methods for classifying data as taught by Kim in the speaker classification system of Sohn in order to more accurately learn unclassified data on the fly (Kim Introduction).
Consider claim 16, Sohn and Khoury teach The artificial intelligence apparatus of claim 1 but do not specifically teach wherein the processor further calculates a density score for the utterance data using the neural network model based on an aggregated winning count of a winner node corresponding to the utterance data, and identifies the speaker of the utterance data as a new speaker when the uncertainty score is greater than or equal to the reference value and the density score is greater than or equal to a density reference value.
In the same field of data classification, Kim teaches wherein the processor further calculates a density score for the utterance data using the neural network model based on an aggregated winning count of a winner node corresponding to the utterance data, and identifies the speaker of the utterance data as a new speaker when the uncertainty score is greater than or equal to the reference value and the density score is greater than or equal to a density reference value (section 4.1, node formation, nodes with highest count T, equivalent to density score, is chosen as the winner. Section 4.1, where no classification nodes activated based on similarity, i.e, below a threshold, new node is created. In combination with Sohn, the labels would correspond to speakers ).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to apply the MPART learning methods for classifying data as taught by Kim in the speaker classification system of Sohn in order to more accurately learn unclassified data on the fly (Kim Introduction).
Claim(s) 13 and 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Sohn and Khoury in view of Kim as applied to claim 12 above, and further in view of Graylin et al. (US PAP 20180359349).
Consider claim 13, Sohn and Khoury and Kim teach the artificial intelligence apparatus of claim 12, but do not specifically teach wherein the processor, when checking whether the response utterance data satisfies the active query, if the response utterance data does not satisfy the active query, re-outputs the active query for the new speaker.
In the same field of user response systems, Graylin teaches wherein the processor, when checking whether the response utterance data satisfies the active query, if the response utterance data does not satisfy the active query, re-outputs the active query for the new speaker ( 0062 if user does not respond, system may repeat the prompt).
It would have been obvious to one of ordinary skill in the art at the time of effective filing to reprompt the user as taught by Graylin the system of Sohn and Khoury and Kim in order to allow for the dialog to be effectively completed to register a new user.
Consider claim 14, Sohn, Kim and Graylin teach the artificial intelligence apparatus of claim 13, wherein the processor, when subsequence response utterance data to the re-outputted active query is input, checks whether the subsequent response utterance data satisfies the active query, and if the subsequent response utterance data does not satisfy the re-outputted active query, unlabels the speaker of the response utterance data (Graylin062 if user does not respond to a reprompt, the system may give up the intent or stop the operation. In combination with Sohn and Kim, the operation would be registering a new speaker, and in Kim the node would not receive a label.).
Allowable Subject Matter
Claims 8-10 would be allowable if rewritten to include all of the limitations of the base claim and any intervening claims. The following is a statement of reasons for the indication of allowable subject matter:
Consider claim 8, Sohn and Khoury and Kim teaches the artificial intelligence apparatus of claim 3, but does not specifically teach “wherein the processor, when identifying the new speaker, collects information of the second node connected to the first node by the edge based on the connection relationship of the edge, calculates an uncertainty score of the first node based on the collected information of the second node, and identifies the speaker of the utterance data as a new speaker if the calculated uncertainty score is greater than or equal to a reference value” when combined with each and every other limitation of the claim, the base claim, and intervening claims. While Kim does discuss measuring uncertainty, Kim does not seem to suggest using it to form new nodes. Therefore claim 8 contains allowable subject matter.
Claims 9 and 10 depend on and further limit claim 8 and therefore contain allowable subject matter as well.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DOUGLAS C GODBOLD whose telephone number is (571)270-1451. The examiner can normally be reached 6:30am-5pm Monday-Thursday.
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, Andrew Flanders can be reached at (571)272-7516. 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.
DOUGLAS GODBOLD
Examiner
Art Unit 2655
/DOUGLAS GODBOLD/Primary Examiner, Art Unit 2655
Prosecution Insights