DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, 18/624,439 filed 04/02/2024 is a Continuation of 17/455,692, filed 11/19/2021, now U.S. Patent 11,995,545, and wherein 17/455,692 is a Continuation of 16/520,899, filed 07/24/2019, now U.S. Patent 11,195,172.
The effective filing date is after the AIA date of March 16, 2013, and so the application is being examined under the “first inventor to file” provisions of the AIA .
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 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.
Status of the Application
This Final Office Action is in response to Applicant’s communication of 09/10/2025.
Claims 1-20 are pending, of which claims 1, 8, and 15 are independent.
All pending claims have been examined on the merits.
Claim Rejections - 35 USC § 103
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
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 set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied 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 non-obviousness.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over S.N. Srihari et al.: "Machine Learning for Signature Verification", in Studies in Computational Intelligence (SCI) 90, pp.387–408 (Year: 2008) henceforth “Srihari”, in view of US 2019/0114743 A1 to Lund et al. (“Lund”. Filed on Dec. 21, 2018. Published on Apr. 18, 2019), and further in view of the Google Patents English Language Translation of FR-2865875-A1 to Donescu (“Donescu”. Eff. Filed on Jan. 29, 2004. Published on Aug. 5, 2005).
In regards to claim 1, Srihari teaches:
1. (Currently Amended) A method, comprising:
…
applying, by the device, transformations to the images of the names to generate a set of images,
wherein an image, of the set of images, comprises a name of the names, and
wherein a transformation, of the transformations, transforms a straight line portion of a character of the name to follow a curve;
…
(See Srihari, p.392: "2.3 Cell Alignment by Spline Mapping
With registration information from the previous stage, a geometric transformation can be constructed to map the landmarks of the reference image– also called control points– to correspondence in the test image. Thin-plate spline warping is a powerful spatial transformation to achieve this goal. An imaginary infinite thin steel plate is constrained to lie over the displacement points and the spline is the superposition of eigenvectors of the bending energy matrix. The spline algebra tries to express the deformation by minimizing the physical bending energy over the flat surface. The resulting plate surface is differentiable everywhere. The spline function ... along x and y coordinates can map every point in the reference image into one in the test image. Please see [26] for details.”)
Thus from the reference grid we can generate the corresponding grid using the mapping function obtained. The aligned cell is defined as the region whose boundaries are the generated grids (Fig. 4).
Fig. 4. Signature grids: (a) using horizontal and vertical pixel histograms, the image is divided by a 4 × 8 grid, (b) mapping grid obtained by spline warping.”)
However, under a conservative interpretation of Srihari, it could be argued that Srihari does not explicitly teach the italicized portions below, which are taught by Lund:
applying, by a device, fonts to names to generate images of the names;
(See Lund, para. [0039]: “As will be described in more detail below, the offline training of the convolutional neural network involves providing document components such as faces, signatures, names addresses, and the like to a simulated document generator, which then generates simulated documents and simulated document images from the document components. The simulated document images and the information from the document components are then used by a neural network learner 352 which implements a learning service to train a convolutional neural network. While the training process implemented by the neural network learner is known in the field, this process conventionally requires vast quantities of actual data. For example, in order to train a convolutional neural network to recognize driver's licenses, it would be necessary to provide large numbers of images of actual driver's licenses to the system. The convolutional neural network would process the images and extract information from the images. This extracted information would then be compared to information that was expected to be extracted, and corrections would be made to the weights of the filters used in the convolution layers of the convolutional neural network in order to improve the network's recognition accuracy moving forward. In many cases, the large amount of pre-existing data (images and expected recognition output) often makes training expensive and impractical (or even impossible). The use of a simulated data generator eliminates the need for a data source that can supply the large numbers of actual document images by generating simulated document images from document components and providing them to the neural network learner.”)
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
receiving, by the device, an image of a signature, wherein a signature in the image of the signature is associated with the name; and
performing, by the device, a transaction based on processing the image of the signature to recognize the name.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
However, under a conservative interpretation of Srihari and Lund, it could be argued that Srihari and Lund does not explicitly teach the italicized portions below, which are taught by Donescu:
wherein the curve is based on a sinusoidal wave;
(See Google Patents English Language Translation of Donescu: “The spatio-frequency transformation is for example a wavelet transformation or a discrete cosine transformation by blocks. Thus, the invention can be applied to images represented in one of the JPEG2000 or JPEG compressed formats.”)
(See Google Patents English Language Translation of Donescu: “In both cases (resolution 1 and resolution 2), according to the JPEG standard, the starting image (in the spatial domain) is divided into blocks, typically of size 8×8, and a discrete cosine transformation (DCT according to the acronym English) is applied on each 8x8 block. Each block contains 64 frequency coefficients corresponding to an 8x8 block of the image in the spatial domain. It is possible to group together sets of blocks into macroblocks as illustrated in FIG. 4b, which makes it possible to obtain a spatio-frequency unit similar to the JPEG2000 standard precinct. The invention applies in this case also.”)
(See Google Patents English Language Translation of Donescu: “Regarding the JPEG format, it is also known that the frequency coefficients in the field of discrete cosine transformation are transposed and change sign in some cases (see US Patent 6,298,166 to Ratnakar et al. , and titled "Image transformations in the compressed domain," Oct. 2001).”)
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, and the “Systems and methods for image modification and image based content capture and extraction in neural networks”, as taught by Lund, with “Signature generation method, involves generating signature of digital image” as taught by Donescu, because the Discrete Cosine Transform (DCT) is a well-known technique for generating a “signature” of an image, that can then be used to train a neural network classifier to quantify how much images are similar or different (e.g. images containing human signatures).
In regards to claim 2, Srihari teaches:
2. (Original) The method of claim 1, further comprising:
identifying an account associated with the name; and
wherein performing the transaction comprises:
performing the transaction associated with the account.
The Examiner interprets that these are examples of intended use. According to MPEP § 707.07(f), form paragraph 7.37.09, “a recitation of the intended use of the claimed invention must result in a structural difference between the claimed invention and the prior art in order to patentably distinguish the claimed invention from the prior art. If the prior art structure is capable of performing the intended use, then it meets the claim.”
In regards to claim 3, Lund teaches:
3. (Original) The method of claim 1,
wherein the processing of the image of the signature is performed using a neural network model.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 4, Lund teaches:
4. (Original) The method of claim 1, wherein applying the transformations to the images of the names comprises:
modifying an orientation of the image based on an angle of rotation.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 5, Lund teaches:
5. (Original) The method of claim 1, wherein applying the transformations to the images of the names comprises:
modifying an intensity of the character to mimic a fading signature.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 6, Lund teaches:
6. (Original) The method of claim 1, wherein the signature is a handwritten signature, and the method further comprising:
verifying that a user associated with the handwritten signature is authorized to conduct the transaction.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
In regards to claim 7, Lund teaches:
7. (Original) The method of claim 1, wherein the name includes a first name and a last name for an individual.
(See also Lund, para. [0094]: “For example, consider a first training data set that includes ten simulated images which are identical except for the names of the fictitious people associated with the images, and a second training data set that includes ten simulated images, each of which has a unique name, address, driver's license number, photograph and image variation (tilt, skew, etc.). Training a neural network with the second set will, in general, enable the neural network to recognize a wider variety of input images than if the neural network were trained with the first data set.”)
The Examiner holds that it is obvious that the names of people in many countries (but not in all countries) consists of “a first name” (a personal name) and “a last name” (a family name).
In regards to claim 8, Srihari teaches:
8. (Currently Amended) A device, comprising:
one or more memories; and
one or more processors, coupled to the one or more memories, configured to:
…
apply transformations to the images of the names to generate a set of images,
wherein an image, of the set of images, comprises a name of the names, and
wherein a transformation, of the transformations, transforms a straight line portion of contiguous characters of the name to follow a curve;
…
(See Srihari, p.392: "2.3 Cell Alignment by Spline Mapping
With registration information from the previous stage, a geometric transformation can be constructed to map the landmarks of the reference image– also called control points– to correspondence in the test image. Thin-plate spline warping is a powerful spatial transformation to achieve this goal. An imaginary infinite thin steel plate is constrained to lie over the displacement points and the spline is the superposition of eigenvectors of the bending energy matrix. The spline algebra tries to express the deformation by minimizing the physical bending energy over the flat surface. The resulting plate surface is differentiable everywhere. The spline function ... along x and y coordinates can map every point in the reference image into one in the test image. Please see [26] for details.”)
Thus from the reference grid we can generate the corresponding grid using the mapping function obtained. The aligned cell is defined as the region whose boundaries are the generated grids (Fig. 4).
Fig. 4. Signature grids: (a) using horizontal and vertical pixel histograms, the image is divided by a 4 × 8 grid, (b) mapping grid obtained by spline warping.”)
However, under a conservative interpretation of Srihari, it could be argued that Srihari does not explicitly teach the italicized portions below, which are taught by Lund:
apply different fonts to names to generate images of the names;
(See Lund, para. [0039]: “As will be described in more detail below, the offline training of the convolutional neural network involves providing document components such as faces, signatures, names addresses, and the like to a simulated document generator, which then generates simulated documents and simulated document images from the document components. The simulated document images and the information from the document components are then used by a neural network learner 352 which implements a learning service to train a convolutional neural network. While the training process implemented by the neural network learner is known in the field, this process conventionally requires vast quantities of actual data. For example, in order to train a convolutional neural network to recognize driver's licenses, it would be necessary to provide large numbers of images of actual driver's licenses to the system. The convolutional neural network would process the images and extract information from the images. This extracted information would then be compared to information that was expected to be extracted, and corrections would be made to the weights of the filters used in the convolution layers of the convolutional neural network in order to improve the network's recognition accuracy moving forward. In many cases, the large amount of pre-existing data (images and expected recognition output) often makes training expensive and impractical (or even impossible). The use of a simulated data generator eliminates the need for a data source that can supply the large numbers of actual document images by generating simulated document images from document components and providing them to the neural network learner.”)
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
receive an image of a signature, wherein a signature in the image of the signature is associated with the name; and
perform a transaction based on processing the image of the signature to recognize the name.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
However, under a conservative interpretation of Srihari and Lund, it could be argued that Srihari and Lund does not explicitly teach the italicized portions below, which are taught by Donescu:
wherein the curve is based on a sinusoidal wave;
(See Google Patents English Language Translation of Donescu: “The spatio-frequency transformation is for example a wavelet transformation or a discrete cosine transformation by blocks. Thus, the invention can be applied to images represented in one of the JPEG2000 or JPEG compressed formats.”)
(See Google Patents English Language Translation of Donescu: “In both cases (resolution 1 and resolution 2), according to the JPEG standard, the starting image (in the spatial domain) is divided into blocks, typically of size 8×8, and a discrete cosine transformation (DCT according to the acronym English) is applied on each 8x8 block. Each block contains 64 frequency coefficients corresponding to an 8x8 block of the image in the spatial domain. It is possible to group together sets of blocks into macroblocks as illustrated in FIG. 4b, which makes it possible to obtain a spatio-frequency unit similar to the JPEG2000 standard precinct. The invention applies in this case also.”)
(See Google Patents English Language Translation of Donescu: “Regarding the JPEG format, it is also known that the frequency coefficients in the field of discrete cosine transformation are transposed and change sign in some cases (see US Patent 6,298,166 to Ratnakar et al. , and titled "Image transformations in the compressed domain," Oct. 2001).”)
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, and the “Systems and methods for image modification and image based content capture and extraction in neural networks”, as taught by Lund, with “Signature generation method, involves generating signature of digital image” as taught by Donescu, because the Discrete Cosine Transform (DCT) is a well-known technique for generating a “signature” of an image, that can then be used to train a neural network classifier to quantify how much images are similar or different (e.g. images containing human signatures).
In regards to claim 9, Lund teaches:
9. (Original) The device of claim 8, wherein the signature is a handwritten signature, and the one or more processors are further configured to:
verify that a user associated with the handwritten signature is authorized to conduct the transaction.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
In regards to claim 10, Lund teaches:
10. (Original) The device of claim 8, wherein the name includes a first name and a last name for an individual.
(See also Lund, para. [0094]: “For example, consider a first training data set that includes ten simulated images which are identical except for the names of the fictitious people associated with the images, and a second training data set that includes ten simulated images, each of which has a unique name, address, driver's license number, photograph and image variation (tilt, skew, etc.). Training a neural network with the second set will, in general, enable the neural network to recognize a wider variety of input images than if the neural network were trained with the first data set.”)
The Examiner holds that it is obvious that the names of people in many countries (but not in all countries) consists of “a first name” (a personal name) and “a last name” (a family name).
In regards to claim 11, Lund teaches:
11. (Original) The device of claim 8, wherein the one or more processors are further configured to:
train a neural network model, with the set of images, to generate a trained neural network model; and
wherein the one or more processors, to perform the transaction, are configured to:
perform the transaction based on an output of the trained neural network model.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 12, Srihari teaches:
12. (Original) The device of claim 8, wherein the one or more processors, to perform the transaction, are configured to:
permit access to another device based on the signature.
The Examiner interprets that these are examples of intended use. According to MPEP § 707.07(f), form paragraph 7.37.09, “a recitation of the intended use of the claimed invention must result in a structural difference between the claimed invention and the prior art in order to patentably distinguish the claimed invention from the prior art. If the prior art structure is capable of performing the intended use, then it meets the claim.”
In regards to claim 13, Srihari teaches:
13. (Original) The device of claim 8, wherein the one or more processors, to perform the transaction, are configured to:
identify an account associated with the name; and
wherein performing the transaction comprises:
performing the transaction using the account.
The Examiner interprets that these are examples of intended use. According to MPEP § 707.07(f), form paragraph 7.37.09, “a recitation of the intended use of the claimed invention must result in a structural difference between the claimed invention and the prior art in order to patentably distinguish the claimed invention from the prior art. If the prior art structure is capable of performing the intended use, then it meets the claim.”
In regards to claim 14, Lund teaches:
14. (Original) The device of claim 8, wherein the one or more processors, to apply the transformations to the images of the names, are configured to:
modify an intensity of the contiguous characters to mimic a fading signature.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 15, Srihari teaches:
15. (Currently Amended) A non-transitory computer-readable medium storing instructions, the instructions comprising:
one or more instructions that, when executed by one or more processors of a device, cause the one or more processors to:
receive information indicating names of individuals;
…
apply transformations to the images of the names to generate a set of images, wherein an image, of the set of images, comprises a name of the names, and wherein a transformation, of the transformations, transforms a straight line portion of a character of the name to follow a curve;
…
(See Srihari, p.392: "2.3 Cell Alignment by Spline Mapping
With registration information from the previous stage, a geometric transformation can be constructed to map the landmarks of the reference image– also called control points– to correspondence in the test image. Thin-plate spline warping is a powerful spatial transformation to achieve this goal. An imaginary infinite thin steel plate is constrained to lie over the displacement points and the spline is the superposition of eigenvectors of the bending energy matrix. The spline algebra tries to express the deformation by minimizing the physical bending energy over the flat surface. The resulting plate surface is differentiable everywhere. The spline function ... along x and y coordinates can map every point in the reference image into one in the test image. Please see [26] for details.”)
Thus from the reference grid we can generate the corresponding grid using the mapping function obtained. The aligned cell is defined as the region whose boundaries are the generated grids (Fig. 4).
Fig. 4. Signature grids: (a) using horizontal and vertical pixel histograms, the image is divided by a 4 × 8 grid, (b) mapping grid obtained by spline warping.”)
However, under a conservative interpretation of Srihari, it could be argued that Srihari does not explicitly teach the italicized portions below, which are taught by Lund:
apply fonts to the names to generate images of the names;
(See Lund, para. [0039]: “As will be described in more detail below, the offline training of the convolutional neural network involves providing document components such as faces, signatures, names addresses, and the like to a simulated document generator, which then generates simulated documents and simulated document images from the document components. The simulated document images and the information from the document components are then used by a neural network learner 352 which implements a learning service to train a convolutional neural network. While the training process implemented by the neural network learner is known in the field, this process conventionally requires vast quantities of actual data. For example, in order to train a convolutional neural network to recognize driver's licenses, it would be necessary to provide large numbers of images of actual driver's licenses to the system. The convolutional neural network would process the images and extract information from the images. This extracted information would then be compared to information that was expected to be extracted, and corrections would be made to the weights of the filters used in the convolution layers of the convolutional neural network in order to improve the network's recognition accuracy moving forward. In many cases, the large amount of pre-existing data (images and expected recognition output) often makes training expensive and impractical (or even impossible). The use of a simulated data generator eliminates the need for a data source that can supply the large numbers of actual document images by generating simulated document images from document components and providing them to the neural network learner.”)
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
receive an image of a signature, wherein a signature in the image of the signature is associated with the name; and
perform a transaction based on processing the image of the signature to recognize the name.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
However, under a conservative interpretation of Srihari and Lund, it could be argued that Srihari and Lund does not explicitly teach the italicized portions below, which are taught by Donescu:
wherein the curve is based on a sinusoidal wave;
(See Google Patents English Language Translation of Donescu: “The spatio-frequency transformation is for example a wavelet transformation or a discrete cosine transformation by blocks. Thus, the invention can be applied to images represented in one of the JPEG2000 or JPEG compressed formats.”)
(See Google Patents English Language Translation of Donescu: “In both cases (resolution 1 and resolution 2), according to the JPEG standard, the starting image (in the spatial domain) is divided into blocks, typically of size 8×8, and a discrete cosine transformation (DCT according to the acronym English) is applied on each 8x8 block. Each block contains 64 frequency coefficients corresponding to an 8x8 block of the image in the spatial domain. It is possible to group together sets of blocks into macroblocks as illustrated in FIG. 4b, which makes it possible to obtain a spatio-frequency unit similar to the JPEG2000 standard precinct. The invention applies in this case also.”)
(See Google Patents English Language Translation of Donescu: “Regarding the JPEG format, it is also known that the frequency coefficients in the field of discrete cosine transformation are transposed and change sign in some cases (see US Patent 6,298,166 to Ratnakar et al. , and titled "Image transformations in the compressed domain," Oct. 2001).”)
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, with “Systems and methods for image modification and image based content capture and extraction in neural networks”, as further taught by Lund, because: “The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.” (See Lund, para. [0084]).
It would have been obvious to a person having ordinary skill in the art (PHOSITA), before the effective filing date of the claimed invention, to include in the method for “Machine Learning for Signature Verification”, as taught by Srihari, and the “Systems and methods for image modification and image based content capture and extraction in neural networks”, as taught by Lund, with “Signature generation method, involves generating signature of digital image” as taught by Donescu, because the Discrete Cosine Transform (DCT) is a well-known technique for generating a “signature” of an image, that can then be used to train a neural network classifier to quantify how much images are similar or different (e.g. images containing human signatures).
In regards to claim 16, Lund teaches:
16. (Original) The non-transitory computer-readable medium of claim 15, wherein the name includes a first name and a last name.
(See also Lund, para. [0094]: “For example, consider a first training data set that includes ten simulated images which are identical except for the names of the fictitious people associated with the images, and a second training data set that includes ten simulated images, each of which has a unique name, address, driver's license number, photograph and image variation (tilt, skew, etc.). Training a neural network with the second set will, in general, enable the neural network to recognize a wider variety of input images than if the neural network were trained with the first data set.”)
The Examiner holds that it is obvious that the names of people in many countries (but not in all countries) consists of “a first name” (a personal name) and “a last name” (a family name).
In regards to claim 17, Lund teaches:
17. (Original) The non-transitory computer-readable medium of claim 15, wherein the signature is a handwritten signature.
(See Lund, para. [0052]: “In some embodiments, the system may be configured to detect security features in the processed image. Many common documents have security features that are designed to allow verification of the authenticity of the document, prevent unauthorized copying of the document, or otherwise secure the document.”)
(See also Lund, para. [0053]: “For instance, in some embodiments, it may be desirable to detect and identify security features in an image in order to verify that the image represents an authentic document. In this case, the convolutional neural network may generate a heat map for the security feature so that the feature can be extracted, identified or otherwise verified by the system.”)
The Examiner notes that traditionally, signatures have been used as “security features” in written documents that were used to verify that the document represents an authentic document.
In regards to claim 18, Lund teaches:
18. (Original) The non-transitory computer-readable medium of claim 15,
wherein the information indicating the names of the individuals is provided in a non- cursive or printed font.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
The Examiner interprets that the disclosure of “The use of different fonts, colors, sizes, etc.” implies the use of non-cursive or printed fonts.
In regards to claim 19, Lund teaches:
19. (Original) The non-transitory computer-readable medium of claim 15, wherein the one or more instructions further cause the one or more processors to:
train a neural network model, with the set of images, to generate a trained neural network model; and
recognize the name using the trained neural network model.
(See also Lund, para. [0084]: “It should be noted that, in addition to changing the simulated document image with image variations such as tilt, skew, centering, lighting, etc., variations may be introduced into the simulated document prior to the generation of the simulated document image through the generation of the simulated document using different fonts, colors and sizes for the document components. The use of different fonts, colors, sizes, etc. in the generation of the simulated documents will provide more variation in the corresponding images so that, when the convolutional neural network is trained with these images, it will help the system to perform even more robust detection.”)
In regards to claim 20,
20. (Original) The non-transitory computer-readable medium of claim 15, wherein the transaction is:
a real estate transaction, or a financial transaction.
The Examiner interprets that these are examples of intended use. According to MPEP § 707.07(f), form paragraph 7.37.09, “a recitation of the intended use of the claimed invention must result in a structural difference between the claimed invention and the prior art in order to patentably distinguish the claimed invention from the prior art. If the prior art structure is capable of performing the intended use, then it meets the claim.”
Response to Arguments
Re: Double Patenting
The Double Patenting rejection was previously withdrawn, in response to Applicant’s filing of a Terminal Disclaimer.
Re: Claim Rejections - 35 USC § 103
The 35 USC § 103 rejection has been amended, as necessitated by Applicant’s amendments to the claims.
Re: Notice of References Cited
The Examiner notes that the article by Eltrabelsi and Lawgali, "Offline Handwritten Signature Recognition based on Discrete Cosine Transform and Artificial Neural Network" is relevant due to its teaching of “Handwritten Signature Recognition based on Discrete Cosine Transform and Artificial Neural Network”, but was published on December 14, 2021 and thus is too recent to qualify as prior art.
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 extension fee 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 date of this final action.
Any inquiry concerning this communication or earlier communications should be directed to Examiner Ayal Sharon, whose telephone number is (571) 272-5614, and fax number is (571) 273-1794. The Examiner can normally be reached from Monday to Friday between 9 AM and 6 PM. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, SPE Christine Behncke can be reached at (571) 272-8103 or at christine.behncke@uspto.gov. The fax 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.
Sincerely,
/Ayal I. Sharon/
Examiner, Art Unit 3695
March 30, 2026