US 10,832,692 B1Granted

Machine learning system for matching groups of related media files

AssigneeAmazon Technologies, Inc.

InventorsRonald Diaz, Srikanth Kotagiri

Life of this patent

Estimating the term…

Loading patent…

Abstract

Techniques are described for verifying that an audio file corresponds to an instance of media content. An audio file is divided into a plurality of audio segments, and a digital fingerprint is generated for each of the plurality of audio segments. A digital signature is generated for the audio file by aggregating the digital fingerprints. The generated digital signature and at least one other digital signature corresponding to an instance of media content are processed as inputs to a linear regression machine learning model, to determine a measure of similarity between the generated digital signature and the at least one other digital signature. The linear regression machine learning model can be trained using a supervised learning approach and a set of training data. Embodiments determine whether the audio file corresponds to the instance of media content, based on the measure of similarity.

Claims20

Dependent claims

A method, comprising:

dividing a first audio file comprising audio in a first language into a plurality of audio segments;
generating a plurality of digital fingerprints by generating a digital fingerprint for each of the plurality of audio segments, wherein frequencies within a predefined range of frequencies are excluded in generating the digital fingerprint, and wherein the predefined range of frequencies generally corresponds to a range of frequencies of human speech;
generating a digital signature for the first audio file by aggregating the plurality of digital fingerprints;
processing the generated digital signature and at least one other digital signature corresponding to an instance of media content as inputs to a linear regression machine learning model, to determine a measure of similarity between the generated digital signature and the at least one other digital signature, wherein the at least one other digital signature is associated with a second audio file comprising audio in a second language and wherein the linear regression machine learning model was trained using a supervised learning approach and a set of training data;
and
determining whether the first audio file comprising the audio in the first language corresponds to the instance of media content comprising a plurality of audio files associated with a plurality of languages, based on the measure of similarity.

3 more elements

A system, comprising:

one or more computer processors;
and
a memory containing computer program code that, when executed by operation of the one or more computer processors, performs an operation comprising:
generating a first digital signature that uniquely represents a first digital audio file comprising audio in a first language;
processing, using a machine learning model, the generated first digital signature and a second digital signature associated with a second digital audio file comprising audio in a second language, to determine a likelihood that the first digital audio file comprising the audio in the first language is associated with an instance of media content comprising a plurality of digital audio files, including the second digital audio file, associated with a plurality of languages;
and
upon determining that the first digital audio file comprising the audio in the first language is not associated with the instance of media content comprising the plurality of digital audio files associated with the plurality of languages, based on the determined likelihood, flagging the first digital audio file.

4 more elements

A non-transitory computer-readable medium containing computer program code that, when executed by operation of one or more computer processors, performs an operation comprising:

generating a first digital signature that uniquely represents a first digital audio file comprising audio in a first language;
processing, using a machine learning model, the generated first digital signature and a second digital signature associated with a second digital audio file comprising audio in a second language, to determine a likelihood that the first digital audio file comprising the audio in the first language is associated with an instance of media content comprising a plurality of digital audio files, including the second digital audio file, associated with a plurality of languages;
and
upon determining that the first digital audio file is not associated with the instance of media content comprising the plurality of digital audio files associated with the plurality of languages, based on the determined likelihood, flagging the first digital audio file.

1 more element

Dependencies are inferred from the claim text.

Facts

Application: 16/049,369Filed 30 Jul 2018
Granted: 10 Nov 2020
Term ends: Not estimatedThe source records don’t support an estimate.

Classification10 codes

G10L 19/018
Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals,…:…
G10L 25/54
For comparison or discrimination: for retrieval
G06F 16/683
Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually: using metadata automatically…
G06N 7/00
Computing arrangements based on specific mathematical models