Speaker Recognition Filtering for False Wake Word Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic speech recognition devices face issues with false activation due to computer-generated voices emitting the personal assistant's name, leading to nonsensical responses and user frustration, especially when multiple devices are involved and lack direct communication.
Innovation Solution
Implement speaker recognition to identify and filter out computer-generated voices, using speaker verification to learn the characteristics of the personal assistant's voice and suppress false triggers, even from devices within hearing range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If keyphrase detection is used to wake the personal assistant, then the assistant can be activated by user voice commands, but false activation occurs when computer-generated voices emit the assistant's name
Solution Approach 1:
The patent introduces speaker recognition technology as an intermediary layer between the audio input and the personal assistant activation. This intermediary analyzes the captured audio signal to determine whether it originates from a human speaker or a computer-generated voice, and only permits activation when a human speaker is detected. This resolves the contradiction by maintaining automatic activation capability while filtering out false activations from synthetic voices.
2Adaptability or versatility
If the personal assistant listens to all audio signals, then it can respond to user commands, but it mistakenly responds to its own emitted keyphrases and other device emissions
Solution Approach 1:
The patent extracts and identifies the harmful component in the audio signal by using speaker recognition to detect computer-generated voices. Once identified, these synthetic voice signals are excluded from triggering the personal assistant. This allows the assistant to maintain its response capability to human users while removing the harmful self-trigger responses to its own emitted keyphrases and other device emissions.
3Productivity
If multiple devices with the same personal assistant are within listening range, then they can all be activated independently, but they cause interference and false triggers to each other
Solution Approach 1:
The patent employs speaker recognition as an intermediary mechanism that enables multiple devices to coexist without interference. Each device analyzes the captured audio signal to verify it originates from a human speaker rather than another device's emitted sound. This intermediary verification layer allows independent activation capability while eliminating inter-device interference and false triggers.
Data Source
AI summary
Techniques related to a method and system of audio false keyphrase rejection using speaker recognition are described herein. Such techniques use speaker recognition of a computer originated voice to omit actions triggered when a keyphrase is present in captured audio and omitted when speech of the captured audio was spoken by the computer originated voice.