Voice-Controlled Device False Wake Word Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Natural language control devices face challenges in accurately distinguishing intended commands from background noise or unintended speech, leading to false positive recognitions and unintended device actions, which can disrupt user experiences.

Innovation Solution

Implementing multiple operating modes with varying recognition settings, including a 'slumber mode' with higher recognition certainty for specific commands, to reduce false positives and adapt to different interaction scenarios, and utilizing wake words to trigger natural language command recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If wake word recognition is enabled to trigger natural language commands, then device responsiveness to user commands is improved, but false positive recognitions from background noise increase

Engineering Contradiction:

Improvedevice responsivenessVSAvoidfalse positive recognitions

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system dynamically adjusts wake word recognition sensitivity parameters based on detected operating conditions. When background noise levels are detected, the system raises the threshold for wake word recognition, reducing false positives while maintaining responsiveness during quiet environments

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The wake word recognition system transitions between different sensitivity states based on environmental context. The system continuously monitors audio characteristics and dynamically switches between high-sensitivity mode (for responsive command recognition) and low-sensitivity mode (for reducing false positives in noisy environments)

Inventive Principle:

Principle #15Dynamics

2Ease of operation

If natural language command recognition is always active, then user interaction convenience is improved, but unintended actions from background speech increase

Engineering Contradiction:

Improveuser interaction convenienceVSAvoidunintended actions

Core Design Contradiction:

Ease of operationVSObject-generated harmful factors

Solution Approach 1:

The wake word serves as an intermediary trigger between background speech and command execution. The system continuously listens for wake words to activate natural language processing, ensuring that only speech intended as commands (preceded by wake word) triggers device actions, while filtering out unintended responses to background conversation

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary wake word detection before activating full natural language command processing. This preliminary check ensures that the subsequent command recognition is triggered only when the user intentionally speaks the wake word, preventing unintended actions from background speech

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS9368105B1Preventing false wake word detections with a voice-controlled device

Publication Date: 2016.06.14 AMAZON TECH INC

AI summary

Natural language controlled devices may be configured to activate command recognition in response to one or more wake words. Techniques are provided to allow for multiple operating modes in which different recognition parameters are employed in recognizing wake words that activate the natural language control functionality of a computing device.