Hotword Speaker Diarization for Privacy-Safe Voice Commands

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In a speech-enabled environment, distinguishing user-directed speech from background noise and unintended speech is challenging, leading to privacy concerns and potential misinterpretation of commands or queries.

Innovation Solution

Implementing a hotword system with speaker diarization to identify user-initiated speech and suppress or obfuscate background speech using a speaker diarization module, which analyzes audio data to separate and process only user-initiated speech.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the system picks up all utterances made in the surrounding environment, then the system can capture all potential user-directed speech, but the system cannot distinguish user-directed speech from background noise and unintended speech, leading to privacy concerns and misinterpretation

Engineering Contradiction:
Improveaccuracy of command interpretationVSAvoidprivacy violation from capturing unintended speech
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent segments the audio stream into distinct speaker portions using speaker diarization technology. This divides the continuous audio input into separate segments attributed to different speakers, allowing the system to process only the segments from the hotword speaker while excluding segments from other speakers, thereby resolving the contradiction between capturing all speech and protecting privacy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts only the relevant portion of the audio stream - specifically the speech from the speaker who uttered the hotword - while removing or suppressing other speakers' speech. This extraction approach allows the system to maintain high reliability in command interpretation while eliminating privacy concerns related to capturing unintended speech from bystanders

Inventive Principle:
Principle #2Taking out (Extraction)

2Loss of information

If the system processes all audio data from all speakers, then the system captures complete audio information, but the system cannot identify which speaker initiated the interaction, leading to misinterpretation of commands

Engineering Contradiction:
Improvecompleteness of audio dataVSAvoidaccuracy of speaker identification
Core Design Contradiction:
Loss of informationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary speaker identification and segmentation before processing the audio data for command interpretation. By using speaker diarization to pre-identify which speaker uttered the hotword and segmenting the audio accordingly, the system ensures accurate speaker attribution before further processing, preventing misinterpretation of commands from the wrong speaker

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3855429B1Speaker diarizaton
Publication Date: 2026.04.15 GOOGLE LLC
  • EP3855429B1 patent drawingFigure 1
  • EP3855429B1 patent drawingFigure 2
  • EP3855429B1 patent drawingFigure 3

AI summary

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speaker diarization are disclosed. In one aspect, a method includes the actions of receiving audio data corresponding to an utterance. The actions further include determining that the audio data includes an utterance of a predefined hotword spoken by a first speaker. The actions further include identifying a first portion of the audio data that includes speech from the first speaker. The actions further include identifying a second portion of the audio data that includes speech from a second, different speaker. The actions further include transmitting the first portion of the audio data that includes speech from the first speaker and suppressing transmission of the second portion of the audio data that includes speech from the second, different speaker.