US Patent Application 18446420. ADAPTING HOTWORD RECOGNITION BASED ON PERSONALIZED NEGATIVES simplified abstract

From WikiPatents
Jump to navigation Jump to search

ADAPTING HOTWORD RECOGNITION BASED ON PERSONALIZED NEGATIVES

Organization Name

GOOGLE LLC

Inventor(s)

Aleksandar Kracun of New York NY (US)

Matthew Sharifi of Kilchberg (CH)

ADAPTING HOTWORD RECOGNITION BASED ON PERSONALIZED NEGATIVES - A simplified explanation of the abstract

This abstract first appeared for US patent application 18446420 titled 'ADAPTING HOTWORD RECOGNITION BASED ON PERSONALIZED NEGATIVES

Simplified Explanation

This patent application describes a method for improving hotword recognition in streaming audio captured by a user device.

  • The method involves receiving audio data that represents a hotword event detected by a first stage hotword detector.
  • A second stage hotword detector then processes the audio data to determine if a hotword is detected in a specific segment of the audio.
  • If the hotword is not detected by the second stage hotword detector, the first segment of the audio data is classified as containing a negative hotword that caused a false detection by the first stage hotword detector.
  • The method then updates the first stage hotword detector based on the classified negative hotword segment, in order to prevent triggering the hotword event in subsequent audio data that contains the negative hotword.

Overall, this patent application presents a method for improving the accuracy of hotword detection by analyzing and classifying segments of audio data, and using this information to update the hotword detection system.


Original Abstract Submitted

A method for adapting hotword recognition includes receiving audio data characterizing a hotword event detected by a first stage hotword detector in streaming audio captured by a user device. The method also includes processing, using a second stage hotword detector, the audio data to determine whether a hotword is detected by the second stage hotword detector in a first segment of the audio data. When the hotword is not detected by the second stage hotword detector, the method includes, classifying the first segment of the audio data as containing a negative hotword that caused a false detection of the hotword event in the streaming audio by the first stage hotword detector. Based on the first segment of the audio data classified as containing the negative hotword, the method includes updating the first stage hotword detector to prevent triggering the hotword event in subsequent audio data that contains the negative hotword.