View source for 18047609. SYSTEM AND METHOD FOR SPEAKER VERIFICATION FOR VOICE ASSISTANT simplified abstract (SAMSUNG ELECTRONICS CO., LTD.)

=SYSTEM AND METHOD FOR SPEAKER VERIFICATION FOR VOICE ASSISTANT=


==Organization Name==

[[:Category:SAMSUNG ELECTRONICS CO., LTD.|SAMSUNG ELECTRONICS CO., LTD.]]


[[Category:SAMSUNG ELECTRONICS CO., LTD.]]

==Inventor(s)==

[[:Category:Myungjong Kim of Milpitas CA (US)|Myungjong Kim of Milpitas CA (US)]][[Category:Myungjong Kim of Milpitas CA (US)]]

[[:Category:Taeyeon Ki of Milpitas CA (US)|Taeyeon Ki of Milpitas CA (US)]][[Category:Taeyeon Ki of Milpitas CA (US)]]

[[:Category:Cindy Sushen Tseng of Santa Clara CA (US)|Cindy Sushen Tseng of Santa Clara CA (US)]][[Category:Cindy Sushen Tseng of Santa Clara CA (US)]]

[[:Category:Srinivasa Rao Ponakala of Sunnyvale CA (US)|Srinivasa Rao Ponakala of Sunnyvale CA (US)]][[Category:Srinivasa Rao Ponakala of Sunnyvale CA (US)]]

[[:Category:Vijendra Raj Apsingekar of San Jose CA (US)|Vijendra Raj Apsingekar of San Jose CA (US)]][[Category:Vijendra Raj Apsingekar of San Jose CA (US)]]

==SYSTEM AND METHOD FOR SPEAKER VERIFICATION FOR VOICE ASSISTANT - A simplified explanation of the abstract==


This abstract first appeared for US patent application 18047609 titled 'SYSTEM AND METHOD FOR SPEAKER VERIFICATION FOR VOICE ASSISTANT

==Simplified Explanation==

The patent application describes a method for speaker identification using a wake word or phrase. Here are the key points:

* The method starts by obtaining audio data and identifying a wake word or phrase in the audio.
* An embedding vector is generated based on the identified wake word or phrase.
* A set of previously-generated vectors representing previous utterances of the wake word or phrase is accessed.
* Clustering is performed on the embedding vector and the set of previously-generated vectors to identify a cluster associated with a speaker.
* A speaker vector associated with the speaker is updated based on the embedding vector.
* A speaker verification model is used to determine a similarity score between the updated speaker vector and the embedding vector.
* Based on the similarity score, it is determined whether the speaker providing the utterance matches the speaker associated with the identified cluster.

== Potential Applications ==
This technology has potential applications in various fields, including:

* Voice-controlled devices: The method can be used to accurately identify the speaker and personalize the user experience on voice-controlled devices such as smart speakers or virtual assistants.
* Security systems: Speaker identification can enhance the security of access control systems by verifying the identity of individuals based on their voice.
* Call center analytics: The method can be used to identify and track speakers in call center recordings, enabling better analysis and monitoring of customer interactions.

== Problems Solved ==
The method addresses the following problems:

* Speaker identification: By using clustering and embedding vectors, the method can accurately identify speakers based on their utterances, even in the presence of noise or variations in speech patterns.
* Personalization: The method allows for personalized user experiences by associating specific speakers with their preferences or settings.
* Security: By verifying the identity of speakers, the method enhances the security of voice-controlled devices and access control systems.

== Benefits ==
The use of this technology offers several benefits:

* Improved user experience: Personalized responses and tailored interactions can be provided to individual speakers, enhancing the overall user experience.
* Enhanced security: By accurately identifying speakers, the method improves the security of voice-controlled devices and access control systems, preventing unauthorized access.
* Efficient call center analytics: The method enables better analysis and monitoring of customer interactions in call centers, leading to improved customer service and support.


==Original Abstract Submitted==

A method includes obtaining audio data and identifying an utterance of a wake word or phrase in the audio data. The method also includes generating an embedding vector based on the utterance from the audio data and accessing a set of previously-generated vectors representing previous utterances of the wake word or phrase. The method further includes performing clustering on the embedding vector and the set of previously-generated vectors to identify a cluster including the embedding vector, where the identified cluster is associated with a speaker. The method also includes updating a speaker vector associated with the speaker based on the embedding vector and determining, using a speaker verification model, a similarity score between the updated speaker vector and the embedding vector. In addition, the method includes determining, based on the similarity score, whether a speaker providing the utterance matches the speaker associated with the identified cluster.



[[Category:G10L15/22]]
[[Category:G10L15/08]]