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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 04/17/2026 has been entered.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 02/24/2026, 04/06/2026, and 04/17/2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Response to Arguments
1. Regarding the rejections under 35 U.S.C. § 103, Applicant’s arguments have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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.
2. Claims 1-6, 8-9, 11-16, and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Hennig et al. (US 2022/0328050 A1, hereinafter Hennig) in view of Pandit et al. (US 2024/0406227 A1, hereinafter Pandit) and further in view of Carroll et al. (US 2024/0127825 A1, hereinafter Carroll).
Regarding claim 1, Hennig discloses A computer-implemented method for detecting machine-based speech in calls (Abstract), the method comprising: obtaining, by a computer (Fig. 6, para. 0106), a plurality of audio speech signals for a caller from a caller device (Fig. 2, “Communication Device 116”; para. 0026 “In some embodiments, in connection with a user account 104 associated with a user, the authenticator component 102 can receive voice data representative of a voice purporting to be the user from an unidentified user 114, for example, via a communication device 116 (e.g., via a communication network with which the authenticator component 102 and communication device 116 can be associated.”) corresponding to a plurality of instances of a repeated audible prompt provided to the caller device (para. 0058 “…the authenticator component 102 can receive a first audio signal comprising first voice data (e.g., representative of a voice) from the unidentified user 114.”; para. 0059 “If the voice verification component 118 determines that the first voice print potentially can be fraudulent …the authenticator component 102, or the service representative or VA associated with the service entity (e.g., as instructed or recommended by the authenticator component 102), can request the unidentified user 114 to repeat the phrase or verbal response (e.g., “I did not quite hear what you said. Can you please repeat the phrase (or the verbal response) again?”), even though the authenticator component 102, or the service representative or VA, understood the phrase or other verbal response presented by the unidentified user 114 the first time. The authenticator component 102 can receive a second audio signal comprising second voice data (e.g., representative of the voice) from the unidentified user 114, wherein the second voice data can comprise the repeating of the phrase or verbal response.”); for each instance of the repeated audible prompt…, extracting, by the computer, a set of acoustic features from the audio speech signal…corresponding to the instance of the repeated audible prompt (para. 0058 “The authenticator component 102 (e.g., employing the voice verification component 118, an AI component 214, and/or a voice print generator component 216) can determine first characteristics of the first voice data and/or generate a first voice print, which can have the first characteristics, based at least in part on the results of analyzing the first audio signal (e.g., in real time or substantially in real time).”; para. 0059 “The authenticator component 102 can receive a second audio signal comprising second voice data (e.g., representative of the voice) from the unidentified user 114, wherein the second voice data can comprise the repeating of the phrase or verbal response. The voice verification component 118 can determine the characteristics of the second voice data based at least in part on the results of analyzing the second voice data.”) using a neural network trained for extracting the set of acoustic features as a feature vector embedding for the input audio signal…for the speech instance (para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate determining or inferring voice characteristics or biometrics associated with voice data representative of voices in audio signals, …”; para. 0066 “Components disclosed herein can employ various classification (explicitly trained (e.g., via training data) as well as implicitly trained (e.g., via observing behavior, preferences, historical information, receiving extrinsic information, and so on)) schemes and/or systems (e.g., support vector machines, neural networks, expert systems, Bayesian belief networks, fuzzy logic, data fusion engines, and so on) in connection with performing automatic and/or determined action in connection with the claimed subject matter.”); determining, by the computer, a first similarity score indicating an acoustic repetition similarity for the plurality of audio speech signals based upon each set of acoustic features extracted from each corresponding audio speech signal…for each speech instance corresponding to each instance of the repeated audible prompt (para. 0059 “The voice verification component 118 also can compare the characteristics of the second voice data to the characteristics of the first voice print (or first voice data of the first audio signal), which can now be a previous voice print (or previous voice data) in relation to the second voice data. Based at least in part on such comparison, the voice verification component 118 can determine a second voice resemblance score associated with the second voice data and the first voice data (and/or can determine the first voice resemblance score with regard to a designated voice print as well).”); and identifying, by the computer, the caller as a fraudulent caller in response to determining that the …similarity score satisfies a …similarity threshold (para. 0059 “If, during the first part of the interaction when the unidentified user 114 first presented the phrase or other verbal response, and during the second part of the interaction when the unidentified user 114 repeated the phrase or other verbal response, the unidentified user 114 was replaying a recording of the voice of the user associated with the user account 104 or using an artificially generated voice that emulates the voice of the user, the second voice resemblance score relating to the first voice data and second voice data typically can be relatively high and satisfy the second threshold voice resemblance score, and accordingly, the voice verification component 118 can determine that the first voice data and second voice data can be fraudulent (e.g., as being a replay of a recording of the voice of the user or an artificially generated voice emulating the voice of the user).”).
Hennig does not specifically disclose:
for each instance of the repeated audible prompt…generating, by the computer, a text representation for an audio speech signal corresponding to the instance of the repeated audible prompt using an automatic speech recognition engine trained to identify a speech instance occurring in an audio waveform of an input audio signal …corresponding to the instance of the repeated audible prompt and generate the text representation using the audio waveform at the speech instance … for the speech instance
determining, by the computer…a second similarity score indicating a textual repetition similarity for the plurality of audio speech signals based upon each text representation generated for each corresponding audio speech signal…for each speech instance corresponding to each instance of the repeated audible prompt;
generating, by the computer, a fused similarity score for the caller based upon the first similarity score and the second similarity score; and [identifying, by the computer, the caller as a fraudulent caller in response to determining that the] fused similarity score [satisfies a] combined similarity threshold.
Pandit teaches for each instance of the repeated prompt (for both a current repeated response and an immediate last response: para. 0115 “Once the caller was done responding to the repetition request, the repetition detector module (e.g., 134f) could compare (i) the caller’s current response to (ii) the immediate last response to determine if the current response is a repetition of the immediate last response.”) generating, by the computer (para. 0009), a text representation for an audio speech signal corresponding to the instance of the repeated audible prompt ((Fig. 2A, 207a ‘Transcribed Audio’ for response to audible prompt (‘Repetition Question’ asked via a call; for example see Table 1))) using an automatic speech recognition engine (para. 0024 “In some embodiments, the first transcription or natural language processing data is generated via a speech recognition or natural language processing operation.”) trained to identify a speech instance occurring in an audio waveform of an input audio signal …corresponding to the instance of the repeated audible prompt and generate the text representation using the audio waveform at the speech instance …for the speech instance(Pandit teaches usage of speech recognition models which identify speech in audio that is trained, e.g. Mozilla deep speech: para. 0077) and…determining, by the computer…a second similarity score indicating a textual repetition similarity for the plurality of audio speech signals based upon each text representation generated for each corresponding audio speech signal…corresponding to each instance of the repeated audible prompt (word overlap feature calculated measuring how similar (how much overlap) the current and previous response have textually: para. 0120 “The evaluated system was configured to calculate the cosine similarity between the current response and the last response…Upon removing stop words and punctuation, the evaluated system was configured to calculate the number of words overlapped (i.e., word overlap feature) between the current response and the last response…”) generating, by the computer, a fused similarity score for the caller based upon the first similarity and the second similarity score (multiple questions can be asked from a question pool measuring similarity (including a repetition question measuring textual similarity (para. 0119-0123), and a name question measuring similarity of a repeated name to audio samples trained to detect the keyword (para. 0113)); the results of these questions can be used in combination to determine whether the caller is fraudulent or real: para. 0070-0072 “To perform the interrogative assessment, and as shown in FIG. 2B, the non-person caller detector can first greet (206) the caller and let the caller know that he/she is talking, e.g., to a virtual assistant… Once the caller has responded to the previous question, the non-person caller detector (e.g., 110) can then ask (210) another question from the question pool. …The non-person caller detector (e.g., 110) can then determine (210) if the response from the caller is appropriate or reasonable for the question asked and, e.g., assign a label (appropriate, not appropriate). The non-person caller detector (e.g., 110) may also assign a confidence score with each label.[0071] The non-person caller detector (e.g., 110) may then can ask (210) another question or make a decision. Based on the provided score or label, the non-person caller detector (e.g., 110) can make an estimation or detection of whether the caller is a human or robo-caller. The number of questions the non-person caller detector (e.g., 110) may ask the caller before making this decision can be static or it can be dynamically established depending on the responses provided by the caller. In some embodiments, the number of questions or the decision to ask additional question may employ a determined confidence score of the classifier label generated by the non-person caller detector (e.g., 110). For example, if the non-person caller detector (e.g., 110) determines a high confidence value indicating, based on the caller's response, that the caller is a human or a robocaller after asking two questions, it can skip asking a third question and direct the caller to the user according to its pre-defined workflow operation. [0072] The non-person caller detector (e.g., 110) can be configured to ask the next question if it is not able to make a decision at any current given time...”); [and identifying, by the computer, the caller as a fraudulent caller in response to determining that the] fused similarity score [satisfies a] combined similarity threshold (para. 0071 “For example, if the non-person caller detector (e.g., 110) determines a high confidence value indicating, based on the caller's response, that the caller is a human or a robocaller after asking two questions, it can skip asking a third question and direct the caller to the user according to its pre-defined workflow operation.”).
Hennig and Pandit are considered to be analogous to the claimed invention as
they both are in the same field of deepfake/robocall detection. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Hennig to incorporate the teachings of Pandit in order to specifically generate a text representation for an audio speech signal corresponding to the instance of the repeated audible prompt using an automatic speech recognition engine trained to identify a speech instance occurring in an audio waveform of an input audio signal and generate the text representation using the audio waveform at the speech instance, and to determine a second similarity score for the plurality of audio speech signals based upon each text representation generated for each corresponding audio speech signal. Doing so would be beneficial, as it is reasonable to expect from a human to repeat a statement by either repeating the exact same statement or saying a semantically similar statement, and thus using the word overlap feature would help to identify fake voice (Pandit, see Table 2, ‘Repetition Question’ row and para. 0116-0117). Further, it would have been obvious to have generated a fused similarity score for the caller based upon a first and second similarity score, and to make a decision to identify a caller as fraudulent in response to determining that the fused similarity score satisfies a combined similarity threshold. Doing so would be beneficial, as individual components/measures can compensate for their individual performance with operations from other components/measures (Pandit, para. 0111), leading to a more robust measure for detecting fraudulent callers than by using any individual measure by itself.
Hennig in view of Pandit does not specifically disclose:
[generating…a text representation…using an automatic speech recognition engine trained to identify a speech instance occurring in an audio waveform of an input audio signal] at a start timestamp and an end timestamp for the speech instance [corresponding to the instance of the repeated audible prompt and generate the text representation using the audio waveform at the speech instance] according to the start timestamp and the end timestamp for the speech instance;
[extracting…a set of acoustic features from the audio speech signal] at the start timestamp and the end timestamp for the speech instance [corresponding to the instance of the repeated audible prompt using a neural network trained for extracting the set of acoustic features as a feature vector embedding for the input audio signal] at the start timestamp and the end timestamp for the speech instance;
[determining…a first similarity score…based upon each set of acoustic features extracted from each corresponding audio speech signal] at the start timestamp and the end timestamp for each speech instance [corresponding to each instance of the repeated audible prompt, and a second similarity score…based upon each text representation generated for each corresponding audio speech signal] at the start timestamp and the end timestamp for each speech instance [corresponding to each instance of the repeated prompt].
Carroll teaches [generating…a text representation…using an automatic speech recognition engine trained to identify a speech instance occurring in an audio waveform of an input audio signal] at a start timestamp and an end timestamp for the speech instance (para. 0114 “In step 316, the authentication server 120 compares text extracted from the voice recording data with the unique string generated earlier in the method. The text is extracted by performing ASR techniques on the voice recording data. The ASR may be performed after receipt of the voice recording data (e.g. immediately prior to step 316). Alternatively, ASR may be performed in real-time during the receipt of the audio stream between steps 312 and 315.”; para. 0107 “The authentication server 120 receives the start of the real-time audio stream at step 312 and may generate a time stamp at this point.”; para. 0109 “Once the authentication server 120 has determined that a stop time has been reached, the authentication server 120 sends an instruction to the user device 110 to stop displaying the unique string at step 314. The stop time may be a predetermined amount of time after the initiation time. For example, it may be a predetermined amount of time after sending the unique string at step 309.”) [corresponding to the instance of the repeated audible prompt and generate the text representation using the audio waveform at the speech instance] according to the start timestamp and the end timestamp for the speech instance (text representation generated from speech collected between start time stamp and stop time: para. 0114 “In step 316, the authentication server 120 compares text extracted from the voice recording data with the unique string generated earlier in the method. The text is extracted by performing ASR techniques on the voice recording data. The ASR may be performed after receipt of the voice recording data (e.g. immediately prior to step 316). Alternatively, ASR may be performed in real-time during the receipt of the audio stream between steps 312 and 315.”);
[extracting…a set of acoustic features from the audio speech signal] at the start timestamp and the end timestamp for the speech instance (voiceprint analysis performed for speech collected between start time stamp and stop time: para. 0116 “At step 318, the authentication server 120 performs voice authentication. For example, the authentication server 120 performs one or more of biometric analysis, synthesis detection and replay detection of the voice recording data, using the techniques and modules previously described. As part of the voice authentication process 318, the authentication server 120 uses the information identifying the registered user received in 304 to retrieve a stored voice print from a stored user profile. The voice authentication may alternatively be performed before step 317 and/or before step 316. As a further alternative, the voice authentication may be performed in real-time during the receipt of the audio stream between steps 312 and 315.”; para. 0107 “The authentication server 120 receives the start of the real-time audio stream at step 312 and may generate a time stamp at this point.”; para. 0109 “Once the authentication server 120 has determined that a stop time has been reached, the authentication server 120 sends an instruction to the user device 110 to stop displaying the unique string at step 314. The stop time may be a predetermined amount of time after the initiation time. For example, it may be a predetermined amount of time after sending the unique string at step 309.”) [corresponding to the instance of the repeated audible prompt using a neural network trained for extracting the set of acoustic features as a feature vector embedding for the input audio signal] at the start timestamp and the end timestamp for the speech instance (voiceprint analysis performed for speech collected between start time stamp and stop time: para. 0116 “At step 318, the authentication server 120 performs voice authentication. For example, the authentication server 120 performs one or more of biometric analysis, synthesis detection and replay detection of the voice recording data, using the techniques and modules previously described. As part of the voice authentication process 318, the authentication server 120 uses the information identifying the registered user received in 304 to retrieve a stored voice print from a stored user profile. The voice authentication may alternatively be performed before step 317 and/or before step 316. As a further alternative, the voice authentication may be performed in real-time during the receipt of the audio stream between steps 312 and 315.”; para. 0107 “The authentication server 120 receives the start of the real-time audio stream at step 312 and may generate a time stamp at this point.”);
[determining…a first similarity score…based upon each set of acoustic features extracted from each corresponding audio speech signal] at the start timestamp and the end timestamp for each speech instance (voiceprint authentication performed using speech collected from start timestamp and stop time: para. 0116 “At step 318, the authentication server 120 performs voice authentication. For example, the authentication server 120 performs one or more of biometric analysis, synthesis detection and replay detection of the voice recording data, using the techniques and modules previously described. As part of the voice authentication process 318, the authentication server 120 uses the information identifying the registered user received in 304 to retrieve a stored voice print from a stored user profile.”; para. 0107 “The authentication server 120 receives the start of the real-time audio stream at step 312 and may generate a time stamp at this point.”; para. 0109 “Once the authentication server 120 has determined that a stop time has been reached, the authentication server 120 sends an instruction to the user device 110 to stop displaying the unique string at step 314. The stop time may be a predetermined amount of time after the initiation time. For example, it may be a predetermined amount of time after sending the unique string at step 309.”) [corresponding to each instance of the repeated audible prompt, and a second similarity score…based upon each text representation generated for each corresponding audio speech signal] at the start timestamp and the end timestamp for each speech instance (textual verification performed for speech collected from start timestamp to stop time: para. 0114 “In step 316, the authentication server 120 compares text extracted from the voice recording data with the unique string generated earlier in the method.”; para. 0107 “The authentication server 120 receives the start of the real-time audio stream at step 312 and may generate a time stamp at this point.”; para. 0109 “Once the authentication server 120 has determined that a stop time has been reached, the authentication server 120 sends an instruction to the user device 110 to stop displaying the unique string at step 314. The stop time may be a predetermined amount of time after the initiation time. For example, it may be a predetermined amount of time after sending the unique string at step 309.”) [corresponding to each instance of the repeated prompt].
Hennig, Pandit, and Carroll are considered to be analogous to the claimed invention as they are in the same field of deepfake/robocall detection. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Hennig in view of Pandit to incorporate the teachings of Carroll in order to specifically generate the text and acoustic features using speech at a start timestamp and an end timestamp for the speech instances, and to determine the first and second similarity scores based on the speech signals at the start timestamps and the end timestamps for each speech instance. Doing so would be beneficial, as recording a user speech only between the specified timestamps taught in Carroll would provide additional security for authorization, as fraudulent speech synthesis requires time to synthesize the unique prompted string, and the timing requirements for collecting speech can be set to be shorter than this time, making it difficult or impossible to synthesis the fraudulent speech in sufficient time (para. 0081).
Regarding claim 2, Hennig in view of Pandit and Carroll discloses transmitting, by the computer, each instance of the repeated audible prompt to the caller device (Hennig, para. 0059 “…the authenticator component 102, or the service representative or VA associated with the service entity (e.g., as instructed or recommended by the authenticator component 102), can request the unidentified user 114 to repeat the phrase or verbal response (e.g., “I did not quite hear what you said. Can you please repeat the phrase (or the verbal response) again?”), even though the authenticator component 102, or the service representative or VA, understood the phrase or other verbal response presented by the unidentified user 114 the first time.”).
Regarding claim 3, Hennig in view of Pandit and Carroll discloses wherein an interactive voice response (IVR) program generates the repeated audible prompt to the caller device (Hennig, para. 0059 “the authenticator component 102, or the service representative or VA associated with the service entity (e.g., as instructed or recommended by the authenticator component 102), can request the unidentified user 114 to repeat the phrase or verbal response (e.g., “I did not quite hear what you said. Can you please repeat the phrase (or the verbal response) again?”), even though the authenticator component 102, or the service representative or VA, understood the phrase or other verbal response presented by the unidentified user 114 the first time.”; para. 0026 “the unidentified user 114 can be interacting with an automated system (e.g., automated voice interface system or interactive voice response (IVR) system) or virtual assistant (VA) component of or associated with the system 100, wherein the automated system or VA component can employ automatically generated voices to speak with users and entities that are engaging with the system 100 and/or associated service entity.”).
Regarding claim 4, Hennig in view of Pandit and Carroll discloses wherein the acoustic features include content of the audio speech signal recognized by a machine-learning architecture executed by the computer (Hennig, para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate determining or inferring voice characteristics or biometrics associated with voice data representative of voices in audio signals, performing voice recognition of voices (e.g., determining or inferring voices) represented in voice data, determining or inferring voice prints (e.g., current voice print, designated voice print, or previous voice print) and associated characteristics, …”; para. 0073 “In some embodiments, the voice recognition component 302 can work in conjunction with the AI component 214 to utilize AI, machine learning, or neural network techniques or algorithms to perform voice or speech recognition on voice data presented in an audio signal to determine the characteristics associated with the voice represented by the voice data.”).
Regarding claim 5, Hennig in view of Pandit and Carroll discloses wherein the acoustic features include low-level acoustic features of the audio speech signal recognized by a machine-learning architecture executed by the computer (Hennig, para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate determining or inferring voice characteristics or biometrics associated with voice data representative of voices in audio signals, performing voice recognition of voices (e.g., determining or inferring voices) represented in voice data, determining or inferring voice prints (e.g., current voice print, designated voice print, or previous voice print) and associated characteristics, …”; para. 0028 “The characteristics can comprise, for example, voice or speech patterns of the voice of the user, tone of the voice, cadence or voice inflections of the voice, speed of speech of the user, a digital signal zero-crossings rate associated with the voice, the physical configuration of the user's mouth, throat, or other physiology of the user when speaking, or other characteristics of or associated with the voice data.”).
Regarding claim 6, Hennig in view of Pandit and Carroll discloses wherein the acoustic features include speech patterns of the caller in the audio speech signal recognized by a machine-learning architecture executed by the computer (Hennig, para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate determining or inferring voice characteristics or biometrics associated with voice data representative of voices in audio signals, performing voice recognition of voices (e.g., determining or inferring voices) represented in voice data, determining or inferring voice prints (e.g., current voice print, designated voice print, or previous voice print) and associated characteristics, …”; para. 0028 “The characteristics can comprise, for example, voice or speech patterns of the voice of the user, tone of the voice, cadence or voice inflections of the voice, speed of speech of the user, a digital signal zero-crossings rate associated with the voice, the physical configuration of the user's mouth, throat, or other physiology of the user when speaking, or other characteristics of or associated with the voice data.”).
Regarding claim 8, Hennig in view of Pandit and Carroll discloses applying, by the computer, a neural network architecture on a first speech signal and a second speech signal of the plurality of audio speech signals to determine the first similarity score between the first speech signal and the second speech signal (Hennig, para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate … determining or inferring voice resemblance scores relating to a comparison of voice prints (e.g., comparison of current voice print to designated voice print or previous voice print)”).
Regarding claim 9, Hennig in view of Pandit and Carroll discloses applying, by the computer, a speaker embedding neural network to extract a speaker embedding using the set of acoustic features from each audio speech signal, and to generate the first similarity score by comparing each speaker embedding (Hennig, para. 0061 “As some examples, the AI component 214 can employ AI, machine learning, and/or neural network techniques, analysis, and algorithms to facilitate determining or inferring voice characteristics or biometrics associated with voice data representative of voices in audio signals, performing voice recognition of voices (e.g., determining or inferring voices) represented in voice data, determining or inferring voice prints (e.g., current voice print, designated voice print, or previous voice print) and associated characteristics, …”; para. 0059 “The voice verification component 118 also can compare the characteristics of the second voice data to the characteristics of the first voice print (or first voice data of the first audio signal), which can now be a previous voice print (or previous voice data) in relation to the second voice data. Based at least in part on such comparison, the voice verification component 118 can determine a second voice resemblance score associated with the second voice data and the first voice data (and/or can determine the first voice resemblance score with regard to a designated voice print as well).”).
Regarding claim 11, claim 11 is a system claim with limitations similar to method claim 1, and is thus rejected under similar rationale.
Additionally, Hennig discloses A system for detecting machine-based speech in calls (Fig. 6) comprising: a computer (Fig. 6, 602) comprising one or more processors (Fig. 6, 604) and configured to (para. 0099-0100).
Regarding claim 12, claim 12 is rejected for analogous reasons to claim 2.
Regarding claim 13, claim 13 is rejected for analogous reasons to claim 3.
Regarding claim 14, claim 14 is rejected for analogous reasons to claim 4.
Regarding claim 15, claim 15 is rejected for analogous reasons to claim 5.
Regarding claim 16, claim 16 is rejected for analogous reasons to claim 6.
Regarding claim 18, claim 18 is rejected for analogous reasons to claim 8.
Regarding claim 19, claim 19 is rejected for analogous reasons to claim 9.
3. Claims 7 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Hennig in view of Pandit and Carroll, and further in view of Keret et al. (US 2023/0082094, hereinafter Keret).
Regarding claim 7, Hennig in view of Pandit and Carroll discloses the step to determine the first similarity score between the first speech signal and the second speech signal (Hennig, para. 0059 “The voice verification component 118 also can compare the characteristics of the second voice data to the characteristics of the first voice print (or first voice data of the first audio signal), which can now be a previous voice print (or previous voice data) in relation to the second voice data. Based at least in part on such comparison, the voice verification component 118 can determine a second voice resemblance score associated with the second voice data and the first voice data (and/or can determine the first voice resemblance score with regard to a designated voice print as well).”).
However, Hennig in view of Pandit and Carroll does not specifically disclose applying, by the computer, a dynamic time-warping (DTW) function on a first speech signal and a second speech signal of the plurality of audio speech signals [to determine the similarity score between the first speech signal and the second speech signal].
Keret teaches applying, by the computer, a dynamic time-warping (DTW) function on a first speech signal and a second speech signal of the plurality of audio speech signals to determine the first similarity score between the first speech signal and the second speech signal (Fig. 4, first and second speech signals 402 and 406; para. 0062 “After sliding dynamic time warping matching 412 is performed, an output 414 may be provided. Output 414 may correspond to a best or highest score in a comparison that is returned with a start and end time in each of call A 402 and call B 406.”).
Hennig, Pandit, Carroll, and Keret are considered to be analogous to the claimed invention as they are all in the same field of detecting fraudulent speech. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Hennig in view of Pandit and Carroll to incorporate the teachings of Keret in order to specifically determine the similarity between a first and second audio signal using a dynamic time warping algorithm. Doing so would be beneficial, as dynamic time warping yields a score which can be used to determine fraudulent speech (para. 0062).
Regarding claim 17, claim 17 is rejected for analogous reasons to claim 7.
4. Claims 10 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Hennig in view of Pandit and Carroll, and further in view of Matthews (US 2022/0269922).
Regarding claim 10, Hennig in view of Pandit and Carroll discloses a determining an indication of the caller as one of fraudulent or genuine (Hennig, para. 0059 “If, during the first part of the interaction when the unidentified user 114 first presented the phrase or other verbal response, and during the second part of the interaction when the unidentified user 114 repeated the phrase or other verbal response, the unidentified user 114 was replaying a recording of the voice of the user associated with the user account 104 or using an artificially generated voice that emulates the voice of the user, the second voice resemblance score relating to the first voice data and second voice data typically can be relatively high and satisfy the second threshold voice resemblance score, and accordingly, the voice verification component 118 can determine that the first voice data and second voice data can be fraudulent (e.g., as being a replay of a recording of the voice of the user or an artificially generated voice emulating the voice of the user).”).
However, Hennig in view of Pandit and Carroll does not specifically disclose:
providing, by the computer, via a user interface, an indication [of the caller classified as one of the fraudulent or genuine.]
Matthews teaches providing, by the computer, via a user interface, an indication (para. 0040 “The example report generator 214 of FIG. 2 generates a report based on the output comparison of the output comparator 212 (e.g., corresponding to whether the media is authentic or a deepfake). The report may be a document and/or a signal. The report generator 214 may include the information related to the media in the report (e.g., the type of media, origin of the media, a timestamp of when the media was output, when the media was created, metadata corresponding to the media, where the media was output or obtained from, etc.) and/or may include the media file itself In some examples, the report generator 214 may cause actions to occur at the processing device 108 in response to a determination that the media is a deepfake. For example, the report generator 214 may cause the media to be stopped, paused, and/or blocked. Additionally, the report generator 214 may display a warning, pop-up, and/or any other audio and/or visual indication for the processing device 108 that the media is a deepfake.””; see also claim 8).
Hennig, Pandit, Carroll, and Matthews are considered to be analogous to the claimed invention as they are all in the same field of deepfake detection. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Hennig in view of Pandit and Carroll to incorporate the teachings of Matthews in order to specifically have the indication of the caller classified as fraudulent or genuine displayed via a user interface. Doing so would beneficial, as this would allow for a user to be notified if deepfake audio is detected, enabling a user to be aware if fraudulent activity is occurring (para. 0002).
Regarding claim 20, claim 20 is rejected for analogous reasons to claim 10.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CODY DOUGLAS HUTCHESON whose telephone number is (703)756-1601. The examiner can normally be reached M-F 8:00AM-5:00PM EST.
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, Pierre-Louis Desir can be reached at (571)-272-7799. 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.
/CODY DOUGLAS HUTCHESON/Examiner, Art Unit 2659
/PIERRE LOUIS DESIR/Supervisory Patent Examiner, Art Unit 2659
Prosecution Insights