# US 2026/0179623 A1 · Methods, processors and systems for detecting synthetic audio content

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0179623 A1
- Document ID: US20260179623A1
- Status: Published application
- Kind: A1
- Application number: 19/423,041
- Filed: 2025-12-17
- Published: 2026-06-25
- Earliest claimed: 2024-12-20
- Assignees: Deezer S.A.
- Applicants: Deezer S.A.
- Inventors: Darius AFCHAR HERAVI MOGHADDAM
- Web page: https://patentgrep.com/patent/US20260179623A1

## CPC

- G10L 17/26: Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
- G10L 17/18: Artificial neural networks; Connectionist approaches
- G10L 25/18: the extracted parameters being spectral information of each sub-band
- G10L 25/30: using neural networks
- G10L 25/51: for comparison or discrimination
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

Methods and processors for detecting a synthetic audio signal is disclosed. The method comprising: acquiring a candidate audio signal including a plurality of audio values; generating a set of patches using the plurality of audio values, a given one from the set of patches including a pre-determined number of audio values from the plurality of audio values, generating a combined patch using the set of patches, a given one value in the combined patch being a combination of values from respective ones from the set of patches; classifying the candidate audio signal as the synthetic audio signal using the combined patch.

## Claims

### Claim 1

A method for detecting a synthetic audio signal, the method executable by a processor, the method comprising:
acquiring a candidate audio signal including a plurality of audio values;
generating a set of patches using the plurality of audio values, a given one from the set of patches including a pre-determined number of audio values from the plurality of audio values,
generating a combined patch using the set of patches,
a given one value in the combined patch being a combination of values from respective ones from the set of patches;
classifying the candidate audio signal as the synthetic audio signal using the combined patch, wherein the classifying comprises:
providing the combined patch to a regression model for generating a regression output, the regression output being indicative of a likelihood that the candidate audio signal is the synthetic audio signal; and
comparing the regression output against a pre-determined threshold for determining that the candidate audio signal as the synthetic audio signal.

### Claim 2 (depends on 1)

The method of claim 1, wherein the set of patches is a sequence of patches, the given one from the sequence of patches including a sub-sequence of audio values from the plurality of audio values, the sub-sequence of audio values including the pre-determined number of sequential audio values.

### Claim 3 (depends on 1)

The method of claim 1, wherein the generating a combined patch comprises:
generating a first combined audio value of the combined patch by combining respective first audio values from the set of patches.

### Claim 4 (depends on 3)

The method of claim 3, wherein the first combined audio value is at least one of an average audio value, a mean audio value, and a median audio value of the respective first audio values from the set of patches.

### Claim 5 (depends on 1)

The method of claim 1, wherein the method further comprises normalizing the combined patch.

### Claim 6 (depends on 1)

The method of claim 1, wherein the regressing model is a Neural Network (NN).

### Claim 7 (depends on 1)

The method of claim 1, wherein the pre-determined number of audio values is a first patch size, and the method further comprises:
generating an other set of patches using the plurality of audio values, a given one from the other set of patches including an other pre-determined number of audio values from the plurality of audio values, the other pre-determined number of audio values being a second patch size
generating an other combined patch using the other set of patches,
a given one value in the other combined patch being a combination of values from respective ones from the other set of patches;
generating a fingerprint using the combined patch and the other combined patch; and
wherein the classifying comprises classifying the candidate audio signal as the synthetic audio signal using the fingerprint.

### Claim 8 (depends on 1)

The method of claim 1, wherein the candidate audio signal is a first candidate audio signal provider by a given audio signal provider, and the combined patch is a first combined patch, the method further comprises:
acquiring a second candidate audio signal including a second plurality of audio values, the second candidate audio signal being originated by the same given audio signal provider;
generating a second set of patches using the second plurality of audio values, a given one from the second set of patches including a second pre-determined number of audio values from the second plurality of audio values,
generating a second combined patch using the second set of patches,
a given one value in the second combined patch being a combination of values from respective ones from the second set of patches; and
classifying the second candidate audio signal as a second synthetic audio signal using the second combined patch.

### Claim 9 (depends on 8)

The method of claim 8, wherein the method further comprises identifying at least one characteristic of an AI model used by the audio signal provider for generating the synthetic audio signal and the second synthetic audio signal based on at least one of the first combined patch and the second combined patch.

### Claim 10

A processor for detecting a synthetic audio signal, the processor being configured to:
acquire a candidate audio signal including a plurality of audio values;
generate a set of patches using the plurality of audio values, a given one from the set of patches including a pre-determined number of audio values from the plurality of audio values,
generate a combined patch using the set of patches,
a given one value in the combined patch being a combination of values from respective ones from the set of patches;
classify the candidate audio signal as the synthetic audio signal using the combined patch, wherein to classify the candidate audio signal, the processor is configured to:
provide the combined patch to a regression model for generating a regression output, the regression output being indicative of a likelihood that the candidate audio signal is the synthetic audio signal; and
compare the regression output against a pre-determined threshold for determining that the candidate audio signal as the synthetic audio signal.

### Claim 11 (depends on 10)

The processor of claim 10, wherein the set of patches is a sequence of patches, the given one from the sequence of patches including a sub-sequence of audio values from the plurality of audio values, the sub-sequence of audio values including the pre-determined number of sequential audio values.

### Claim 12 (depends on 10)

The processor of claim 10, wherein to generating a combined patch comprises the processor configured to:
generate a first combined audio value of the combined patch by combining respective first audio values from the set of patches.

### Claim 13 (depends on 12)

The processor of claim 12, wherein the first combined audio value is at least one of an average audio value, a mean audio value, and a median audio value of the respective first audio values from the set of patches.

### Claim 14 (depends on 10)

The processor of claim 10, wherein the processor is further configured to normalize the combined patch.

### Claim 15 (depends on 10)

The processor of claim 10, wherein the regressing model is a Neural Network (NN).

### Claim 16 (depends on 10)

The processor of claim 10, wherein the pre-determined number of audio values is a first patch size, and the processor is further configured to:
generate an other set of patches using the plurality of audio values, a given one from the other set of patches including an other pre-determined number of audio values from the plurality of audio values, the other pre-determined number of audio values being a second patch size generate an other combined patch using the other set of patches,
a given one value in the other combined patch being a combination of values from respective ones from the other set of patches;
generate a fingerprint using the combined patch and the other combined patch; and
wherein the to classify, the processor is configured to classify the candidate audio signal as the synthetic audio signal using the fingerprint.

### Claim 17 (depends on 10)

The processor of claim 10, wherein the candidate audio signal is a first candidate audio signal provider by a given audio signal provider, and the combined patch is a first combined patch, the processor being further configured to:
acquire a second candidate audio signal including a second plurality of audio values, the second candidate audio signal being originated by the same given audio signal provider;
generate a second set of patches using the second plurality of audio values, a given one from the second set of patches including a second pre-determined number of audio values from the second plurality of audio values,
generate a second combined patch using the second set of patches,
a given one value in the second combined patch being a combination of values from respective ones from the second set of patches; and
classify the second candidate audio signal as a second synthetic audio signal using the second combined patch.

### Claim 18

A non-transitory computer-readable medium comprising executable instructions which, when executed by at least one processor, cause the at least one processor carry out steps of a method for detecting a synthetic audio signal, the method executable by a processor, the method comprising:
acquiring a candidate audio signal including a plurality of audio values;
generating a set of patches using the plurality of audio values, a given one from the set of patches including a pre-determined number of audio values from the plurality of audio values,
generating a combined patch using the set of patches,
a given one value in the combined patch being a combination of values from respective ones from the set of patches;
classifying the candidate audio signal as the synthetic audio signal using the combined patch, wherein the classifying comprises:
providing the combined patch to a regression model for generating a regression output, the regression output being indicative of a likelihood that the candidate audio signal is the synthetic audio signal; and
comparing the regression output against a pre-determined threshold for determining that the candidate audio signal as the synthetic audio signal.

### Claim 19 (depends on 18)

The non-transitory computer-readable medium of claim 18, wherein the set of patches is a sequence of patches, the given one from the sequence of patches including a sub-sequence of audio values from the plurality of audio values, the sub-sequence of audio values including the pre-determined number of sequential audio values.

### Claim 20 (depends on 18)

The non-transitory computer-readable medium of claim 18, wherein the generating a combined patch comprises:
generating a first combined audio value of the combined patch by combining respective first audio values from the set of patches.
