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 .
All objections/rejections not mentioned in this Office Action have been withdrawn by the Examiner.
Information Disclosure Statement
The information disclosure statement(s) (IDS) submitted on 14 May 2026 is/are being considered by the examiner.
Status of the Claims
Prior to entry of the amendment(s) and/or consideration of the argument(s), the status of the claims is as follows.
Claim(s) 1-6 is/are pending.
Claim(s) 1-6 are rejected under 35 U.S.C. §101 because the claimed invention is directed to an abstract idea without significantly more.
Claim(s) 1, 3-6 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Ahmed (U.S. Pat. App. Pub. No. 2023/0282202, hereinafter Ahmed).
Claim(s) 2 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ahmed as applied to claim 1 above, and further in view of Non-patent literature to Alwahab (Alwahab, D.A., Zaghar, D.R. and Laki, S., 2018, July. FIR filter design based neural network. In 2018 11th International Symposium on Communication Systems, Networks & Digital Signal Processing (CSNDSP) (pp. 1-4). IEEE.(2018), hereinafter Alwahab).
Response to Amendments
Applicant’s amendment filed on 03 March 2026 has been entered.
In view of the amendment to the claim(s), the amendment of claim(s) 4; the cancellation of claim(s) 1-3 and 5-6; and the addition of claim(s) 7-12 have been acknowledged and entered.
In view of the amendment to claim(s) 4 and the cancellation of claim(s) 2-3 and 5, the objection to claim(s) 2-5 is withdrawn.
In view of the amendment of claim(s) 4 and the cancellation of claim(s) 1-3 and 5-6, the rejection of claims 1-6 under 35 U.S.C. §102 and 103 is withdrawn.
In light of the amended/newly added claims, new grounds for rejection under 35 U.S.C. §103 and 35 U.S.C. §112 are provided in the action below.
Response to Arguments
Applicant’s arguments regarding the subject matter rejections under 35 U.S.C. §101 and prior art rejections under 35 U.S.C. §102/103, see pages 7-22 of the Response to Non-Final Office Action dated 04 December 2025, which was received on 03 March 2026 (hereinafter Response and Office Action, respectively), have been fully considered.
Applicant’s arguments regarding the analysis under 35 U.S.C. §101 as performed in the Office Action is duly noted. However, said arguments are not persuasive. The claims which were subject to the rejection under 35 U.S.C. §101 are no longer present. Though applicant has amended claim 4, amended claim 4 now crosses two statutory classes and contains significant ambiguity as to what active steps described with reference to the preamble devices are included in the described method (specifically, the method steps described with relation to components of the method, which are now incorporated into the preamble as part of the audio data discrimination device) and which ones are cited merely as prefatory language. As the present claims do not reflect the claims as previously presented, the rejection of claims 1-6 is withdrawn.
With respect to the rejection(s) of claim(s) 4 under 35 U.S.C. §102(a)(2) as being anticipated by Ahmed, applicant asserts that Ahmed fails to teach or suggest the “unique combination and arrangement of features recited in Applicant’s claim 4.” (Response, pgs. 5-6) Applicant’s arguments are addressed individually below.
Initially, applicant sets out what appears to be a list of secondary considerations, in pages 10-15 of the Response, indicating that the applicant’s claims reflect improvements in the art, focusing primarily on the asserted advantages of a CPU-based system. However, such considerations are moot in the context of rejections under 35 U.S.C. §102, as the relevant question for secondary considerations is whether the claims are obvious or not, and has no bearing on anticipation.
Regarding the general arguments of deficiency, applicant's arguments fail to comply with 37 CFR 1.111(b) because they amount to a general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references. At pgs. 16-17 of the Response, applicant further asserts that claim 4 is amended to recite the features of “an up-sampling unit that obtains up-sampled multi-stream data by performing up-sampling processing on each of the plurality of stream data,” “an audio data discrimination device including: a global feature discriminator that includes a learnable function unit and discriminates authenticity of audio data based on a plurality of global features of audio data, the global features each being obtained from the audio data in a range defined by each of a plurality of scales each of which defines a different range, from each other, to be determined for the audio data; and a detailed feature discriminator that includes a learnable function unit and discriminates authenticity of audio data based on a plurality of detailed features of audio data, the detailed features each being obtained by acquiring the audio data in each of a plurality of periods, which are different from each other, and converting the acquired data of each period into two-dimensional data,” “using the processor to perform a generator parameter updating step of updating parameters of the convolution processing unit of the audio data generation device and parameters of the learnable function unit of the multi-stream generation unit based on the loss evaluation data obtained in the loss evaluation step,” and “using the processor to perform a discriminator parameter updating step of, based on the loss evaluation data obtained in the loss evaluation step, updating the parameters of the learnable function unit of the global feature discriminator of the audio data discrimination device, and updating parameters of the learnable function unit of the discriminator of the detailed feature discriminator of the audio data discrimination device.” Applicant then asserts that “Ahmed et al. does not teach or suggest the unique combination and arrangement of features recited in Applicant's claim 4, including the above-emphasized features.” (Response, pg. 17).
As best as can be understood by the Office, the above listing references a laundry-list of amendments and then, without describing a specific deficiency, conclusorily states that Ahmed fails to teach or suggest all “features… including the above-emphasized features,” where this conclusion is somehow related to the “unique combination and arrangement,” as described by the applicant. (Id.)
Such arguments fail to specifically enumerate how the cited portions of the Ahmed fail to teach or suggest each individual cited limitation, and, as such, amount to no more than attorney argument. Therefore, these arguments are insufficient to amount to a proper traversal of the rejection and the rejection is maintained in light of said arguments.
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., “performing MRF processing on mel-spectrogram data”; “achieved only using a CPU”) are not recited in the rejected claim(s), either as originally presented or as amended. Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993). It is noted that applicant specifically provides analysis regarding asserted differences in upsampling and the GAN generator of Ahmed. (Response, pgs. 18-19). However, applicant provides no direct connection between any specific limitation of claim 4, either as previously presented or as currently amended, and the argued deficiencies. The asserted deficiencies of Ahmed are related to claim 4 only based on the non-sequitur that Ahmed “clearly fails to teach or suggest the unique combination and arrangement of features…including…” the same laundry-list assertion of limitations which applicant believes are not taught by Ahmed (and fail to comply with 37 CFR 1.111(b), as explained above). Since the specific limitations argued are not present in claim 4 and have no clear association to the cited limitations of the laundry-list of assertedly deficient components, said arguments fail to properly traverse the rejection. Applicant is invited to amend the claims, during normal prosecution, such that the claims reflect the desired limitations and said limitations can be properly considered.
Regarding applicants analysis as provided in support of the arguments, applicant’s interpretation of Ahmed is duly noted and will be taken under advisement. However, such analysis is not considered binding regarding either the reference itself or other art generally. Further, it is important to note that applicant mischaracterizes the disclosure of Ahmed regarding at least the processing requirements. Applicant specifically argues that “[t]o perform high-performance adversarial learning with such a model, the generator must use GPUs (as discussed in paragraphs [0038], [0115], and [0116] of Ahmed et al.), and the diffusion model in the GAN generator utilizes a GPU.” (Response, pg. 18). From this, applicant concludes that one skilled in the art would have understood that this method is not achievable “using only a CPU.” (Id.)
However, though asserted by the applicant that “the generator must use GPUs,” Ahmed both doesn’t support the conclusion and states the exact opposite of applicant’s conclusion. Paragraph [0038] is a background section paragraph which (1) describes prior art GAN vocoders (MeIGAN and Parallel WaveGAN) during inference from which Ahmed is attempting to distinguish, and (2) even if this was directed to .
At paragraphs [0115] and [0116], Ahmed explains that “StyleMeIGAN is a fully convolutional, feed-forward model… [which] allows for highly parallelizable generation, several times faster than real time on both [central] processing units, CPUs, and graphic processing units, GPUs.” Examiner is not aware of a valid interpretation of these paragraphs which would support the conclusion that the GAN generator described in Ahmed “must use GPUs,” as any such interpretation would have to rationally explain how “StyleMeIGAN… allows for highly parallelizable generation, several times faster than real time” on “CPUs”, somehow results in a system which cannot use CPUs and “must use GPUs.”
The most charitable interpretation of applicant’s assertion that these paragraphs somehow support that Ahmed “must use GPUs” relies on treating a clear clerical error in Ahmed as intentional and assuming that “control” in “control processing unit” isn’t just a typographical error. Even if the “control processing unit” were treated as intentional, the best which could be argued is that Ahmed was merely silent regarding the use of “central processing units.” However, since the acronym CPU is well known in the art as “central processing unit”, there is no standard recognized computing component called a “control processing unit”, and given that the CPU vs. GPU paradigm is a foundational architecture consideration for machine learning, one having ordinary skill in the art would immediately recognize CPU as referring to “central processing unit and that the word “control” is a trivial spelling error. Therefore, applicant’s interpretation is not persuasive. Even if applicant’s described CPU limitation, as indicated above, were recited in claim 4 as described above, such a limitation would likely be rejectable in light of paragraphs [0115]-[0116] of Ahmed. However, full consideration of such an amendment is reserved for when said limitation is actually presented in the claims.
Regarding claims 7-12, applicant treats these claims as dependent from claim 4 and argues that these claims are “allowable for at least the same reasons that claim 4 is allowable.” However, this argument is not persuasive.
Applicant is advised regarding the dependency of the claims 7-12. The Federal Circuit has indicated that “A claim's status as dependent or independent depends on the substance of the claim in light of the language of § 112, ¶ 4, and not the form alone.” (Monsanto v. Syngenta Seeds, 503 F.3d 1352, 1357 (Fed. Cir. 2007)). The court further explained that, under pre-AIA 35 U.S.C § 112, ¶ 4 (currently 35 U.S.C § 112(d)), “[t]o establish whether a claim is dependent upon another, this court examines if the new claim both refers to an earlier claim and further limits that referent.” (Monsanto at 1357, citing 35 U.S.C § 112, ¶ 4 (2000)). Specifically with reference to dependence from a process claim, the court held that whether a claim is properly read as dependent or independent turns on whether the asserted dependent claim “specifically requires …the performance of the steps” in the referent process claim. (Id. at 1358).
During prosecution, it falls to the examiner to determine status as dependent or independent with reference to 35 U.S.C § 112(d). (See MPEP 2173.05(f)). As such, we perform the same analysis here for claims 4 and 7-12.
The listed independent claim appears to correspond to claim 4 as this is the only claim which is clearly in independent form, contain no statements indicating reliance on another claim, and claim 4 is expressly indicated as the rationale for allowability for claims 7-12. The remaining claims, claims 7-12 are understood, as indicated by the applicant, to be asserted as dependent claims.
Regarding claims 8-9, these claims both refer to claim 7 and provide further limitation to that claim. As well, under the broadest reasonable interpretation, claims 8-9 specifically require the performance of the steps in the referent process claim. As such, though not dependent from claim 4, claims 8-9 are understood as dependent claims of claim 7 for the purposes of further prosecution.
With reference to the substance of claims 7, and 10-12, these claims “refer to an earlier claim” but the fail to “further limit that referent,” as the limitations are directed to changing the statutory class of the referent claim. Claim 4 recites a process which “consist[s] of a series of steps or acts to be performed.” (MPEP 2106). Claims 7 and 10-12 are asserted as dependent from claim 4, but act only to convert one or more portions of the previously cited process and/or one or more portions of the audio data generation device which applicant extracts from that asserted process, to another statutory class. Each of claims 7, and 10-12 is discussed individually below.
Regarding claim 7, claim 7 is directed to “an audio data generation device,” in which applicant both ignores the established existence of the “audio data generation device” in claim 4 and then, as part of a limitation in the body of the claim, recites “when the audio data generation device is used to perform the adversarial learning method for the audio data generation device according to claim 4,” which under the BRI appears to be either (1) a conditional limitation which purports to exchange the “audio data generation device” already existing in claim 4 with the newly generated “audio data generation device” of claim 7 for performing the adversarial learning method of claim 4; or (2) a statement regarding a selected result from the action of the existing “audio data generation device” of claim 4. Due to the lack of antecedent basis between the two components, as explained further in the 112(b) rejection below, which of the two is correct is unclear. Claim 7 references claim 4 a total of three (3) more times, indicating various changes/incorporations of specific components of claim 4 as presented.
As a result, the apparent a la carte claim format for a device claim, which (as best understood in light of numerous conflicting claim limitations) appears to further alter preamble device limitations of the independent method claim, is only directed to specific method limitations in the method claim and/or functional limitations as described with reference to the device. Therefore, claim 7 does not necessarily require the performance of all limitations of claim 4, including the limitations which are incorporated in the preamble, and claim 7 is understood as merely incorporating limitations from claim 4 and is not a dependent claim of claim 4.
Regarding claim 10, claim 10 describes a learning method, which is apparently distinguished from the “adversarial learning method” of claim 4. This learning method purports to incorporate the “audio data generation device” described in the preamble of adversarial learning method of claim 4, and otherwise does not incorporate the remaining parts of claim 4. As such, claim 10 does not necessarily require the performance of all limitations of claim 4, claim 10 is understood as merely incorporating limitations from claim 4 and is not a dependent claim of claim 4.
Regarding claim 11, claim 11 recites a “speech synthesis processing system” which incorporate the “audio data generation device” of claim 7. However, as the “audio data generation device” incorporates method limitations from claim 4 in a conditional manner, which are not understood as part of the device itself. To further explain, claim 7, being conditioned on whether or not a specific method step is performed, is not a configuration of the device such that incorporation of the device in a system is necessarily understood to incorporate said limitations. Therefore, the incorporated limitations of the method of claim 4, as incorporated into the device of claim 7, are not necessarily incorporated into the system of claim 11. For at least these reasons, claim 11 is understood as merely incorporating limitations from claim 7 and is not a dependent of claims 4 or 7.
Regarding claim 12, claim 12 is directed to “an audio data generation device,” in which applicant both ignores the established existence of the “audio data generation device” in claim 4 and then, as part of a limitation in the body of the claim, recites “when learning processing of the adversarial learning method according to claim 4 has been completed,” which under the BRI appears to be a modification of an unrelated audio data generation device using the results of the process described in claim 4. Due to the lack of antecedent basis between the two components, as explained further in the 112(b) rejection below, there is no way to know whether this is the device described in the preamble of claim 4 or a separate device. However, as claim 12 only requires “first optimal parameters” and “second optimal parameters,” where the applicant merely defines said parameters in the context of claim 4, indicating “optimal parameters” as “being parameters that have been set in the multi-stream generation unit when learning processing of the adversarial learning method according to claim 4 has been completed.” As the “audio data generation device” of claim 12 is not the audio data generation device of claim 4, The mere existence of parameters which share a similarity to those generated by the process of claim 4, regardless how they are actually derived, does not necessarily require the performance of all limitations of claim 4. For at least these reasons, claim 12 is understood as merely incorporating limitations from claim 4 and is not a dependent of claims 4.
Therefore, claims 7, 10, 11, and 12 are not understood as a result of, a continuation of, or a further limit to the process of claim 4. As such, the amended claim set includes five (5) independent claims and a total of seven (7) claims.
If the applicant wishes for these claims to be treated as dependent, applicant is advised to amend the claims, in light of specification support, such that the claims further limit the referenced independent claim.
Of note, in applicant’s previous two claim sets, as presented to the Office in each of applicant’s Patent Application Fee Determination Records (SB06), filed on 25 July 2024 and on 03 March 2026, applicant asserted a number of independent claims and dependent claims which did not correspond to the number of independent claims and dependent claims which were currently pending in the application at the time of response.
As such, applicant is advised against presenting group arguments which traverse the rejection based on dependency from a previously discussed claim. Though all arguments will be given due consideration when presented, applicant’s reliance on dependency in traversing a rejection may result in improper traversal and/or a non-responsive submission, depending on the specific circumstances of the provided response.
The Applicant has not provided any further statement and therefore, the Examiner directs the Applicant to the below rationale.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 4, and 7-12 are rejected under 35 U.S.C. §112(b) or 35 U.S.C. §112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 4, it is unclear whether the active steps performed by the “audio data generation device” are considered part of the “adversarial learning method,” thus rendering the claim indefinite. Claim 4 describes an “adversarial learning method to be performed using an audio data generation device” including “a multi-stream generation unit that… obtains multiple stream data from mel spectrogram data”; “an up-sampling unit that… obtains up-sampled multi-stream data” and includes “performing up-sampling processing on each of the plurality of stream data”; “a convolution processing unit… that obtains audio waveform data” and includes “performing convolution processing on the up-sampled multi-stream data”; “a global feature discriminator that… discriminates authenticity of audio data based on a plurality of global features of audio data”; and “a detailed feature discriminator that… discriminates authenticity of audio data based on a plurality of detailed features of audio data.” Each of these limitations occur in the preamble of claim 4 as amended, and indicate specific elements which are currently performed by the device, and only after these elements are disclosed does the claim assert that the method is “being performed with a processor and a memory that the processor can access,” and provides specifically enumerated method limitations.
Regarding specific limitations directed to the devices in the preamble, their patentable weight turns on whether said limitations are actively performed in the method or not. For example, the further functional limitations added by the applicant to the audio data discrimination device, defining global features and detailed features are presented in the preamble and further defining how each are obtained, have no direct application to the method steps and do not describe structural limitations for any component of the audio data discrimination device. It is noted that the only limitation which relies on the “audio data discrimination device” is “a discrimination step of inputting audio data generated by the audio data generation device or reference audio data into the audio data discrimination device, and causing the audio data discrimination device to discriminate authenticity of the input data.” (Claim 4, lines 23-25). With regards to limitations to the device in the preamble, functional limitations which do not imbue structure to this preamble device would not be entitled to patentable weight, unless such limitations are actually part of the method in some way. The discrimination step is disclosed as “causing the audio data discrimination device to discriminate authenticity of the input data,” which only incorporates the “audio data discrimination device” and components/functionality which are necessary to perform the described discrimination. It is further noted that as the “audio data discrimination device” includes “a global feature discriminator” and “a detailed feature discriminator” but it is not limited to such. As a result, an audio data discrimination device which includes these discriminators, but which actually discriminates using another system or discriminate using these discriminators but in another way would teach these limitations. In essence, all that is required by claim 4 as amended, is discrimination by the discrimination device, where the discrimination device has “a global feature discriminator” and “a detailed feature discriminator”. The amended limitations regarding “global features” and “detailed features” describe how they are obtained, thus how the device would be used. As the global features and detailed features are merely a basis for the discriminator and not, in and of themselves incorporated into the device, these limitations are not understood to limit the structure of the device itself in any sense. Since it is not clear whether said active steps described with reference to the audio data discrimination device should be considered part of the method steps or not, one skilled in the art would not know if the embodiments disclosed regarding such features, described in the preamble but otherwise covering apparently active steps, are performed in the recited method or not. Therefore, claim 4 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Further regarding claim 4, the phrase “audio data generated by the audio data generation device or reference audio data” lacks clarity. Claim 4 recites “audio data generated by the audio data generation device or reference audio data” at lines 23-24. However, and as explained above, it is at this point where the entirety of the utility of the preamble audio data generation device lies. In the alternative where reference audio data is used, the “audio data generation device” is not necessarily used in any portion of claim 4. As such, claim 4 recites a limitation where “the audio data generation device” may be superfluous depending on the whether the output of the audio data generation device is provided to the audio data discrimination device or not. As such, one skilled in the art would not know the metes and bounds of the “audio data generation device” as relevant to the “adversarial learning method.” Does the “adversarial learning method” include an “audio data generation device” which “obtains multiple stream data from mel spectrogram data;” “obtains up-sampled multi-stream data by performing up-sampling processing on each of the plurality of stream data;” and “obtains audio waveform data by performing convolution processing on the up-sampled multi-stream data,” if the discrimination step is performed with the “reference audio data”? Said limitations relate to possible outputs during operation. These limitations do not provide or otherwise affect the structure of the “audio data generation device” and, in the case of using reference audio data, could not even be putatively argued to result in a tangible effect to the method as claimed. Are any of those steps performed even when the “audio data” is “generated by the audio data generation device”. The preamble “audio data generation device,” though relying on audio data generation in the name, results in the generation of “audio waveform data,” not “audio data”, and further fails to recite how the “audio data,” which may be used by the method, is actually created. Therefore and for this further reason, claim 4 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Regarding claim 7, the relationship between the “audio data generation device” of claim 7 and the “audio data generation device” of claim 4 lacks clarity. Claim 4 is directed to “an adversarial learning method to be performed using an audio data generation device.” It is unclear what exactly is being added or amended when the applicant purports to “further limit” claim 4 with an audio data generation device, which does not reference or give consideration to the already existing “audio data generation device” of claim 4. Is this entire structure described in claim 7 viewed by the applicant as a further modification of the existing “audio data generation device”? A new device? In either case, there are numerous conflicting parts which lack appropriate antecedent basis and such a claim, even if clarified, would remain an improper dependent claim under 112(d) as it would not further limit the independent claim. Further, this problem is only exacerbated when consideration is given for respective parts which actually perform the method of claim 4, such as the processor and the memory, which are also duplicated in claim 7, but also without a clear connection to the cited parts and performing completely different and separate functions. Not only do we not know which, if any, of the method steps which are described in claim 4 would be incorporated at each of four (4) locations where portions of claim 4 are purportedly incorporated, but we also do not know whether the active steps described in the preamble would be incorporated for the same reasons as described with reference to claim 4 above. As currently presented, the “audio data generation device” of claim 7 describes borrowing/replacing/incorporating/using components and process results as selected from claim 4, however one hardly knows where to begin in determining exactly what is being borrowed/replaced/incorporated/changed. Therefore, claim 7 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Regarding claims 8 and 9, claims 8 and 9 depend from claim 7 and incorporate all limitations therein. therefore, claims 8 and 9 are rejected as being indefinite under 35 U.S.C. §112(b) for at least the same reasons as claim 7.
Regarding claim 10, the limitation “…for the audio generation device according to claim 4” is unclear. Claim 4 is, purportedly, a method claim. Applicant, in claim 10, appears to be extracting the preamble device of said method claim, but not completely, as this is merely “for the…device” and not using the device. Further, the preamble recites that claim 10 is “a learning method,” which, though referencing the adversarial learning method of claim 4, appears to be distinguished from the “adversarial learning method.” Applicant provides no further relationship connection to the “adversarial learning method” (such as the phrase “further comprising” or others which are well known to patent practitioners), a person having ordinary skill in the art is left to guess whether “the method comprising” is intended to be a continuation of the method described in claim 4, in light of the asserted dependency from the preamble device of claim 4, or not. This, of course, carries forward the lack of clarity described above for claim 4 regarding the method claim which heavily relies on active limitations to the same device described in the preamble but avoids clearly incorporating the same. Therefore, claim 10 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Regarding claim 11, the phrase “the audio data generation device according to claim 7” is unclear. The system of claim 11 incorporates the audio data generation device of claim 7, which may or may not actually be the audio data generation device described in the preamble of method claim 4. At the current level of dependency and interrelationship, with numerous tiers of lack of clarity, alongside no less than two changes of statutory type of patentable subject matter, it is impossible to know what is actually incorporated in claim 11 by the phrase “the audio data generation device according to claim 7.” Therefore, claim 11 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Regarding claim 12, the phrases “audio data generation device,” “first optimal parameters,” and the “second optimal parameters” each lack clarity. Claim 12 recites “an audio data generation device” at line 1. However, as claim 12, similar to claim 7, fails to clarify the relationship between the “audio data generation device” of claim 4, the method claim from which claim 12 borrows/modifies one or more steps/results. As such, and similar to claim 7, claim 12 lacks clarity regarding the apparently related devices described in both claims 4 and 12.
Claim 12 further recites “the first optimal parameters being parameters that have been set in the multi-stream generation unit when learning processing of the adversarial learning method according to claim 4 has been completed” and “the second optimal parameters being parameters that have been set in the convolution processing unit when the learning processing of the adversarial learning method according to claim 4 has been completed.” However, each of these are intervening steps in the method which is not necessarily performed in this claim, for a device which may or may not be the “audio data generation device” described in claim 4. Assuming this is a separate device, the above would not necessarily be the result of the method step but some tangential use of that method step. Which multi-stream generation unit is being modified by the adversarial learning method? Similarly, which convolution processing unit is being modified by the adversarial learning method? The one recited in claim 4 or the one recited in claim 12? Further, what does “have been set…” mean in the context of specific steps performed? Claim 4 does not describe “set[ting]” any of the parameters and the word “set” is not used in claim 4. Though we might guess that “set” is intended to refer to the updating step, such a conclusion would be just that: a guess. As such, claim 12 purports to incorporate portions of claim 4 without clearly indicating what exact portions of the adversarial learning method of claim 4 the applicant is incorporating by reference. Therefore, claim 12 lacks clarity and is rejected as being indefinite under 35 U.S.C. §112(b).
Appropriate correction is required.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 4 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Ahmed (U.S. Pat. App. Pub. No. 2023/0282202, hereinafter Ahmed).
Regarding claim 4, Ahmed discloses An adversarial learning method to be performed using an audio data generation device including (The systems and methods described with reference to “the audio generator 10” including the limitations recited with reference to claim 1 above, and which includes “GAN generator 11” incorporating the waveform synthesis block 1120 “GAN discriminator 100” where “the GAN generator 11 and the GAN discriminator 100 may concur in constituting the audio generator 10” and the GAN discriminator 100 can be used for training of the audio generator. The GAN generator 11 “generate[s] an audio signal” and the GAN discriminator 100 tries to determine whether the “generated audio signal is real... or fake.”; Ahmed, ¶ [0210], [0220]): a multi-stream generation unit (The “audio generator 10” includes a first processing block where “the first processing block 50 may be instantiated by each of a plurality of blocks (in FIG. 1, blocks 50 a, 50 b, 50 c, 50 d, 50 e, 50 f, 50 g, 50 h).”; Ahmed, ¶ [0212]) that includes a learnable function unit (“in the first processing block 40, 50, a conditioning set of learnable layers (e.g., 71, 72, 73) {a learnable function unit} may be used to process the target data 12 and/or the input signal 14.”; Ahmed, ¶ [0212]) and obtains multiple stream data from mel spectrogram data (the first processing block outputs the “first output data 69... in a plurality of channels” as derived from the target data, which can be “a spectrogram (e.g., a mel-spectrogram)” (e.g., “each block 50 a-50 h may be conditioned by the target data 12...such as a mel-spectrogram.”); Ahmed, ¶ [0211], [0213], [0233]); an up-sampling unit that obtains up-sampled multi-stream data by performing up-sampling processing on each of the plurality of stream data (“At each block 30 and 50 a-50 h, the signal, evolving from noise 14 towards becoming speech 16, may be upsampled. For example, at the upsampling block 30 before the first block 50a among the blocks 50 a-50 h, an 88-times upsampling is performed.”; Ahmed, ¶ [0215]); and a convolution processing unit, which is capable of learning parameters for determining convolution processing, that obtains audio waveform data by performing convolution processing on the up-sampled multi-stream data (“At a second processing block 45 (FIGS. 1 and 6), only one single channel may be obtained, and multiple samples are obtained in one single dimension. As can be seen, another TADEResBlock 42 (further to blocks 50 a-50 h) is used (which reduces to one single channel). Then, a convolution layer 44 and an activation function (which may be TanH 46, for example) may be performed” where “conditioning feature parameters 74, 75” as part of TADEResBlock 42 and convolutional layer 44, “may be obtained, e.g. by convolution, during training” and where the processing of the upsampled data at the convolution layer 44 results in “the speech 16” being “obtained (and, possibly, stored, rendered, encoded, etc.)”; Ahmed, ¶ [0212], [0234], [0249], [0309]); and an audio data discrimination device including: a global feature discriminator that includes a learnable function unit and discriminates authenticity of audio data based on a plurality of global features of audio data (“The GAN discriminator 100” and components thereof, such as the described full band discriminator, “shall reduce its own discriminatory loss (e.g., through the method of gradients or other methods) and update its own internal parameters” the full band discriminator being one of “discriminator (132 a-132 d)” which “analyze[s] the sub-bands 120 of the input speech signal (104 or 16)” including, in one or more examples, a “1...subband... adversarial evaluation”; Ahmed, ¶ [0220], [0310]), the global features each being obtained from the audio data in a range defined by each of a plurality of scales each of which defines a different range, from each other, to be determined for the audio data (Applicant is describing intended use for the global features of the global feature discriminator, incorporated into the preamble of the claim. These global features are not being necessarily obtained or used in the method of claim 4, as the global feature discriminator itself is not necessary for the audio data discrimination device to function. Specifically, the method claim only requires “causing the audio data discrimination device to discriminate authenticity of the input data”. Any discrimination of authenticity is sufficient to meet the requirements of that limitation. The preamble of claim 4 describes “an audio data discrimination device including... a global feature discriminator” (emphasis added). The intended use of the global feature discriminator does not change the structure of the audio data discrimination device and is not directly included in or necessarily relied on for the performance of the claimed method. As such, these limitations are not entitled to patentable weight.); and a detailed feature discriminator that includes a learnable function unit and discriminates authenticity of audio data based on a plurality of detailed features of audio data (“The GAN discriminator 100” and components thereof, such as the described multi-subband discriminator, “shall reduce its own discriminatory loss (e.g., through the method of gradients or other methods) and update its own internal parameters” the multi subband discriminator being one of “discriminator (132 a-132 d)” which “analyze[s] the sub-bands 120 of the input speech signal (104 or 16)” including, in one or more examples, a “4 and 8...subband... adversarial evaluation” and “More precisely we may use... 4, and 8 subbands” which “enables a multi-resolution adversarial evaluation of the speech signal (104 or 16) in both time and frequency domains” and allows the system/method “to analyze different frequency bands of the speech signal (104 or 16) during training”; Ahmed, ¶ [0220], [0310]), the detailed features each being obtained by acquiring the audio data in each of a plurality of periods, which are different from each other, and converting the acquired data of each period into two- dimensional data (Applicant is describing intended use for the global features of the detailed feature discriminator, incorporated into the preamble of the claim. These detailed features are not being necessarily obtained or used in the method of claim 4, as the detailed feature discriminator itself is not necessary for the audio data discrimination device to function. Specifically, the method claim only requires “causing the audio data discrimination device to discriminate authenticity of the input data”. Any discrimination of authenticity is sufficient to meet the requirements of that limitation. The preamble of claim 4 describes “an audio data discrimination device including... a global feature discriminator” (emphasis added). The intended use of the global feature discriminator does not change the structure of the audio data discrimination device and is not directly included in or necessarily relied on for the performance of the claimed method. As such, these limitations are not entitled to patentable weight.), the method being performed with a processor and a memory that the processor can access (“Some or all of the method steps may be executed by (or using) a hardware apparatus, like for example, a microprocessor, a programmable computer or an electronic circuit” and “can be performed using a digital storage medium” which is accessible to the processor in a computing device.; Ahmed, ¶ [0345]-[0346]), the method comprising: a discrimination step of inputting audio data generated by the audio data generation device or reference audio data into the audio data discrimination device (“the GAN discriminator 100 may be input by both audio signal 16 synthesized generated by the GAN generator 10, and real audio signal (e.g., real speech) 104 acquired e.g., through a microphone”; Ahmed, ¶ [0220], [0277]), and causing the audio data discrimination device to discriminate authenticity of the input data (Using the input audio signals, the discriminator can “recognize whether the generated audio signal is real... or fake”; Ahmed, ¶ [0220], [0277]); using the processor to perform a loss evaluation step of obtaining loss evaluation data using a loss function based on resultant data of the discrimination step (“During training, a loss function (adversarial loss) 140 may be optimized” where the adversarial loss is obtained by “evaluating a representation of the generated audio signal (16) or a representation of the reference audio signal (104) by one or more evaluators (132)” which may be performed using the processor, where the system can “process the signals to obtain a metric (e.g., loss) which is to be minimized” using the loss function at paragraph [0288], as calculated using D(x) and D(G(z)) which are the resultant data of the discrimination step.; Ahmed, ¶ [0277], [0284], [0287]-[0293], [0345]); using the processor to perform a generator parameter updating step of updating parameters of the convolution processing unit of the audio data generation device and parameters of the learnable function unit of the multi-stream generation unit based on the loss evaluation data obtained in the loss evaluation step (“The GAN generator 11 shall minimize the losses (e.g., through the method of the gradients or other methods), and update the conditioning features parameters 74, 75 by taking into account the results at the GAN discriminator 100” where “Weight normalization may be applied to all convolution operations in... GAN generator 11” and “discriminator 100” and where the loss is minimized based on the “results at the GAN discriminator 100” which may be performed using the processor; Ahmed, ¶ [0220], [0313], [0345]); and using the processor to perform a discriminator parameter updating step of, based on the loss evaluation data obtained in the loss evaluation step, updating the parameters of the learnable function unit of the global feature discriminator of the audio data discrimination device (“The GAN discriminator 100 shall reduce its own discriminatory loss (e.g., through the method of gradients or other methods) and update its own internal parameters” and can “use FB-RWDs with the Adam optimizer with a discriminator learning rate (Ird)” which may be performed using the processor; Ahmed, ¶ [0220], [0313], [0345]), and updating parameters of the learnable function unit of the discriminator of the detailed feature discriminator of the audio data discrimination device. (As indicated above, “discriminator 100” reduces “its own discriminatory loss” which is the loss based on the ability to “recognize real signals 16” as distinguished “from the fake audio signals generated by the GAN generator 11”; Ahmed, ¶ [0220], [0313]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Ahmed (U.S. Pat. App. Pub. No. 2022/0059107) discloses methods, apparatuses and a system for hybrid adversarial-parametric speech synthesis to improve the synthesis of original speech signals using compact-learned-parametric representation by implementing a Generator trained in a Generative Adversarial Network setting in combination with linear predictive coding.
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 Sean E. Serraguard whose telephone number is (313)446-6627. The examiner can normally be reached 07:00-17:00 M-F.
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, Daniel C. Washburn can be reached at (571) 272-5551. 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.
/Sean E Serraguard/Primary Examiner, Art Unit 2657
Prosecution Insights