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 the filing on 02/03/2026 with RCE which entered the amendment filed 01/14/2026. Claims 1-6 are currently pending in the filing of 02/03/2026. Applicant's previous response filed 07/10/2025 also had claims 1-6 pending. Claims 1-6 were also pending in the applicants previous response / amendment filed on 07/10/2025.
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 02/03/2026 has been entered.
Response to Applicant’s Amendments / Arguments Regarding 35 U.S.C. § 103
The applicant’s remarks, on pages 6-9 of the response / amendment, the applicant argues the features which allegedly distinguish over the previously cited references cited in the 35 U.S.C. § 103 rejections.
Applicant’s arguments have been considered but are moot in view of the new ground(s) of rejection.
The examiner would like the thank the applicant’s attorney for the discussion during the interview conducted on 01/14/2026, which the examiner found helpful. The examiner encourages the applicant to call the examiner via telephone after reviewing this action to conduct a brief conversation and/or schedule an interview at any time for the purpose of identifying potentially allowable subject matter.
Previous Claim Objections
The applicant’s response / amendment filed on 02/03/2026 appropriately corrected the claim objections to claim 5 by correctly renumbering claim 5 to depend from claim 4. The previous objections have been withdrawn.
Claim Rejections - 35 USC § 103
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.
Claims 1-3 and 6 are rejected under 35 U.S.C. 103 as being unpatentable over US 10685008 to Kurve et al. (hereinafter Kurve), in view of US 20180039861 to Saito (hereinafter Saito), US 20180027004 to Huang et al. (hereinafter Huang).
Regarding claim 1, Kurve teaches,
A user verifying method by a security token combined with biometric information processing techniques, which is performed by a server, the method comprising: (Abstract and Col. 8, lines 1-3.)
receiving an enrollment request signal including biometric information for enrollment of a user from a user terminal; (col. 1, lines 25-33, teaches enrollment of user’s acoustic features.)
extracting feature information from the biometric information of a plurality of other persons; (fig. 5 and Col. 23, lines 55-67 teaches extraction unit 510 and feature extraction unit 515. Fig. 5 teaches multiple users 505.)
embedding the feature information extracted from the biometric information for enrollment of the user and the biometric information of the plurality of other persons; (Applicant’s printed publication at [0082] describes embedding as converting feature information into binary form.)
(Col 1, lines 20-30 teach enrollment. Col. 2, line 50 teaches use of voiceprints / biometrics. Col. 14, lines 5-15 teaches embedding to maximize locality using Support Vector Machines (SVM) using hyperplanes. See also Col. 18, lines 50-67 describing figs. 4a-4C and the feature selection using SVM.)
generating a projection parameter and a decision hyperplane parameter from the hyperplanes; (Applicant’s printed publication at fig. 2 & [0095] describes examples of the parameters as: “the projection parameter includes a random matrix (Ω) and a random vector (r) as performed in Random Fourier Features (RFF) algorithm discussed below, and [0095] also describes the decision hyperplane parameter includes a vector (w) and a scalar value (b), as performed by SVM as discussed below.)
(Fig. 14B teaches a Random Fourier Features (RFF) algorithm for an RBF kernel, using a random matrix. Col. 17, lines 36-45 teach the random matrix. Col. 13, line 44 to Col. 14, lines 24 teaches explicit kernels that include the use of RBF kernel and/or SVMs. Col. 14, lines 5-14 teaches a kernel using linear SVM and hyperplanes during embedding, by using linear decisions boundaries that include support vectors of a hyperplane.) (See also: Col. 17, lines 34 to 63 teaches RFF being used for RBF kernel. Col. 18, lines 36-38 teaches RFF storing a large random matrix. Col. 8 lines 15 to 30 also teaches SVM choosing hyperparameters.)
storing the projection parameter and the decision hyperplane parameter; (Col. 18, lines 36-38 teaches RFF storing a large random matrix. Col. 17, lines 36-35 also teach the random matrix of fig. 14B. Col. 14, lines 5-15 teach the SVM that learns and stores boundaries that support a hyperplane.)
storing the generated first template; (Col. 18, lines 36-38 teaches RFF storing a large random matrix. Col. 14, lines 5-15 teach the SVM that learns and stores boundaries that support a hyperplane.)
receiving a verification request signal including biometric information for verification of the user from a terminal; (Col. 19, line 1, teaches user authentication)
extracting feature information from the biometric information for verification of the user; (Col. 19, lines 11-16 teaches capturing the user voice biometrics. Fig. 5, col. 23, line 61-62 teaches feature extraction unit 515. See also classification unit.)
embedding the feature information extracted from the biometric information for verification of the user; (Col. 23, lines 44-51. Fig. 5, col. 26, lines 1-13, teaches the output 555 is based on kernel, explicit kernel map.)
comparing the first template with the second template thereby verifying the biometric information for verification of the user against the biometric information for enrollment of the user. (Col. 23, lines 40-54 teaches classifying data points as genuine or fraudulent. See also fig. 5, output 555.)
Kurve fails to teach using a hash function that is used for generating the template,
However, Saito teaches,
generating a first template by applying a cryptographic hashing function to the first key; (fig. 9, hashing process 913 and [0068-70] teach the hashing using hyperplanes. [0125] teaches substituting Locality Sensitivity Hashing in place of normality sensitivity hashing NSH.) (Please also see applicant’s printed publication [0107] discussing hyperplanes.) (The examiner notes that US 20160267637, which is not relied upon in the rejection, at [0100] teaches that locality-sensitive hash (LSH) uses random hyper-plane sampled from a zero-mean multivariate Gaussian.)
generating a second key based on the embedding result for the biometric information for verification of the user, and the projection parameter and the decision hyperplane parameter; (see above discussion regarding fig. 9 and [0068-70] and applicant’s printed publication [0107] discussing hyperplanes.)
generating a second template by applying the cryptographic hashing function to the second key; and (see above discussion regarding fig. 9 and [0068-70] and applicant’s printed publication [0107] discussing hyperplanes.)
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the teachings of Kurve, which teaches using Random Fourier methods, with SVM / kernels to create a database of features that are used in authentication / comparison of biometrics (fig. 5 & Col 23 line 55 to Col. 24 line 25), with Saito, which also teaches training linear support vector machines SVMs ([0003] & [0113]) and using kernels ([0123]), and additionally teaches the use of hashing of hyperplane parameters to create and index / key that may be used to represent data in different multi-plane dimensional data (figs. 9-10). One of ordinary skill in the art would have been motivated to perform such an addition to provide Kurve with the added ability to use hash functions for indexing and/or secure storage of data that has been classified, as taught by Saito, for the purpose of increasing security by storing hash information of biometrics instead of raw biometric information or biometric features / vectors.
Kurve and Saito fail to explicitly teach using a key / parameter to determine the number of hyperplanes / classes,
However, Huang teaches,
generating a first key having a random binary string of a predetermined number of bits; ([0080] teaches local sensitivity hashing (LSH) that uses random hyperplanes. One of skill in the art understands that in Local Sensitivity Hashing (LSH), particularly when using random projections for cosine similarity (e.g., SimHash), hash buckets are directly related to and created by hyperplanes. [0086-91] teaches that the key value mapping where the keys are bucket identifiers \ hyperplanes.)
determining a number of hyperplanes equal to the predetermined number of bits of the first key, (figs. 7a&b, [0080] & [0086-91] teach the hyperplanes / buckets. See also [0080-91].) wherein each hyperplane is determined such that, when a support vector machine (SVM) is applied to the embedding result, an output of the SVM corresponds to a value of a respective bit of the first key; ([0080] teaches comparing the buckets of local sensitivity hashes. See also [0080-91] and [0098].)
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the teachings of Kurve, which teaches using Random Fourier methods, with SVM / kernels to create a database of features that are used in authentication / comparison of biometrics (fig. 5 & Col 23 line 55 to Col. 24 line 25), with Saito, which also teaches training linear support vector machines SVMs ([0003] & [0113]) and using kernels ([0123]), and additionally teaches the use of hashing of hyperplane parameters to create and index / key that may be used to represent data in different multi-plane dimensional data (figs. 9-10), with Huang, which also teaches support vector machines / SVMs and machine learning ([0049]), and additionally teaches the use of local sensitivity hashing / LSH for performing comparisons on data to detect differences ([0080-91]). One of ordinary skill in the art would have been motivated to perform such an addition to provide Kurve and Saito with the added ability to utilize local sensitivity hashing, as taught by Huang, for the purpose of increasing security by hashing data and storing data, so that the data is efficiently stored in a smaller space for efficiency while increasing security by using one-way functions / hashes instead of storing raw data and allowing for comparison of hashes that are locally sensitive.
Regarding claim 2, Kurve, Saito, and Huang teach,
The user verifying method of claim 1, wherein the generating of the projection parameter and the decision hyperplane parameter includes:
Kurve teaches,
performing high-dimensional mapping by applying a random Fourier feature mapping (RFFM) (Random Fourier Feature Mapping (RFFM) is taught by fig. 14B teaches a Random Fourier Features (RFF) algorithm for a kernel.) to the result of embedding the feature information extracted from the biometric information for the enrollment and the biometric information of the plurality of other persons; (fig. 14B and Col 17 line 33 to Col. 18, line 8 teach the RFF that uses the random partitions / grids to construct bit strings in the different dimensions, which is input into the RBF kernel / SMV, discussed below. Col. 19, line 55 teaches training several thousand user profiles.)
determining the hyperplanes by applying a Linear Support Vector Machine (SVM) to the mapping result; and (fig. 14b teaches the RFF is input into the RBF kernel, where RBF is known in the art as a form of SVM. RBF kernel maps are used to transform the feature space to one where the structural embeddings are arranged, as discussed in Col. 14, lines 5-14.) (The examiner also interprets the hyperplanes as a single hyperplane.) (See also Huang, [0049] teaching machine learning and SMVs.)
generating the projection parameter based on the mapping result (RFF of fig. 14b teaches mapping using random Fourier features (Col. 7, lines 4-6) and uses randomness in the matrix and random binning features to random partitions / grids, as discussed in Col 17, lines 37-66.) and generating the decision hyperplane parameter based on the determination result. (Col. 14, lines 5-14 teaches using (RBF) kernels to learn linear decision boundaries using SVM to support hyperplanes.) (Applicant’s printed publication at [0095] describes the projection parameter includes a random matrix (Ω) and a random vector (r).)
Regarding claim 3, Kurve, Saito, and Huang teach,
The user verifying method of claim 1, wherein the extracting of the feature information from the biometric information for the enrollment and the biometric information of the plurality of other persons includes:
padding the feature information extracted from the biometric information for the enrollment and the biometric information of the plurality of other persons into an input space. (Kurve, Col. 19, line 55 teaches training several thousand user profiles. The examiner interprets this feature as adding biometric information during enrollment to the already added biometric information of the other users.) (Applicant’s printed publication at [0080-83] describes “padding” as adding biometric data to the model during training / enrollment by embedding, including embed enrolling biometric along with embedding the other users information. [0093] teaches different subsets of data / subsets of users. [0097] teaches FEHash generating a classifier that finds the user from the padded info (user’s embedded info, and another user’s embedded info).)
Regarding claim 6, Kurve, Saito, and Huang teach,
A computer program stored on a non-transitory computer readable medium, the computer program comprising computer-executable instructions for causing the computer to (Kurve, col. 28, lines 40-50 teach a non-transitory computer readable medium.)
… perform the user verifying methods by a security token combined with biometric information processing techniques of claim 1. (See rejection of claim 1 above.)
Claims 4-5 are rejected under 35 U.S.C. 103 as being unpatentable over Kurve, in view of Saito, in view of Huang, in view of US 20200202181 to Yadav et al. (hereinafter Yadav).
Regarding claim 4, Kurve, Saito, and Huang teach,
The user verifying method of claim 1, wherein the embedding of the feature information extracted from the biometric information for the enrollment and the biometric information of the plurality of other persons includes:
Kurve, Saito, and Huang fail to explicitly teach selecting person from among plurality of other persons,
However, Yadav teaches,
selecting ‘q’ persons among the plurality of other persons based on a length (m) of the first key (s); ([0043] teaches a label (key) that is used to determine positive or negative distances between SVM hyperplanes.) (Applicant’s printed publication at [0093] teaches that different subsets of people are classified differently, with q and p other persons.)
allocating a binary label identical to a binary label of a user to the selected ‘q’ persons; and ([0043] teaches a label (key) that is used to determine positive or negative distances between SVM hyperplanes.)
allocating a binary label, which are not identical to the binary label of the user, to the remaining persons (‘p-q’ persons) selected among the plurality of other persons. (The examiner asserts that the example of other users all having negative distances, would result in a different binary label, when the label of the user has a positive distance.)
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the teachings of Kurve, which teaches using Random Fourier methods, with SVM / kernels to create a database of features that are used in authentication / comparison of biometrics (fig. 5 & Col 23 line 55 to Col. 24 line 25), with Saito, which also teaches training linear support vector machines SVMs ([0003] & [0113]) and using kernels ([0123]), and additionally teaches the use of hashing of hyperplane parameters to create and index / key that may be used to represent data in different multi-plane dimensional data (figs. 9-10), with Huang, which also teaches support vector machines / SVMs and machine learning ([0049]), and additionally teaches the use of local sensitivity hashing / LSH for performing comparisons on data to detect differences ([0080-91]), with Yadav, which also teaches training linear SVMs and use of hyperplanes in the SVMs classification (Abstract), and additionally teaches the use of labels for positive and negative distances between hyperplanes ([0041-44]). One of ordinary skill in the art would have been motivated to perform such an addition to provide Kurve, Saito, and Huang with the added ability to use different labels / keys for information that is classified differently by the SVM based on the positive or negative distance from a hyperplane, as taught by Yadav, for the purpose of increasing efficiency when storing labels (keys) for different objects being classified in security applications.
Regarding claim 5, Kurve, Saito, Huang, and Yadav teach,
The user verifying method of claim 4, wherein the selecting of the ‘q’ persons includes:
selecting the ‘q’ persons according to an equation below:
Yadav teaches,
pCq>=m (Examiner asserts that the situation where there are 2 other people (p), and only the user in the group q, would result in 2 times 1 times 1 times 1 = 2 (p X C X q = 2) when the combination is set to 1, which results in 2 => m, when m = 1, where 1 bit is needed to label the groups as having positive or negative distances between the SVM hyperplanes, as taught in [0043])
(Here, ‘C’ means a combination; ‘p’ is the number of other persons; ‘q’ is the number of persons selected among the plurality of other persons; and ‘m’ is the length of the first key (s)). (see above. [0045] teaches combination.)
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN WILLIAM AVERY whose telephone number is (571) 272-3942. The examiner can normally be reached on 9AM-5PM.
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, Farid Homayounmehr can be reached on (571) 272-3739.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see https://ppair-my.uspto.gov/pair/PrivatePair. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/B.W.A./
/FARID HOMAYOUNMEHR/Supervisory Patent Examiner, Art Unit 2495