Server False Wake Word Detection in Voice Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice-enabled devices sometimes incorrectly activate in response to sounds resembling wake words, which are present in the audio content being played, leading to unintended pauses or volume changes in media playback.
Innovation Solution
Implementing a server-based false wake word detection system that analyzes audio streams for wake words and generates metadata indicating the presence of false wake words, which is then used by voice-enabled devices to deactivate during such instances, preventing incorrect activation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voice-enabled devices continuously monitor audio for wake words, then responsiveness to user commands is improved, but false activations occur when wake words appear in media content
Solution Approach 1:
The system divides audio monitoring into two separate channels: media audio analysis and wake word detection. The media playback audio stream and the wake word detection audio stream are processed independently, allowing the device to identify wake words in media content without triggering false activations. This segmentation resolves the contradiction by maintaining reliability in wake word detection while eliminating false activations from media content.
Solution Approach 2:
The patent introduces an intermediary mechanism where wake word detections from the media audio stream are intercepted and handled separately from user-facing wake word commands. When a wake word is detected in media content, the system identifies it as a false wake word and prevents it from triggering device activation. This intermediary layer resolves the contradiction by filtering out false activations while preserving genuine user commands.
2Ease of operation
If voice-enabled devices pause or adjust volume upon detecting wake words, then command responsiveness is improved, but media playback continuity is disrupted by false wake words
Solution Approach 1:
The system segments wake word handling into two distinct pathways: genuine user commands that trigger device responses, and false wake words from media content that are suppressed. By analyzing the source of the wake word detection through separate audio streams, the device can maintain media playback continuity while still responding to actual user commands. This resolves the contradiction by preserving media duration while maintaining operational responsiveness.
Solution Approach 2:
The system implements feedback mechanisms where wake word detections are analyzed in context to determine their origin. When wake words are detected in the media audio stream, the system provides feedback to suppress the activation and maintain playback continuity. This feedback loop resolves the contradiction by ensuring that only genuine user commands interrupt media playback, while false wake words from the media itself do not cause disruptions.
Data Source
AI summary
A wake word detector, at a server of a content delivery network (CDN) that provides audio (or other) content to a device, such as a voice-enabled device, detects false wake words in the audio content. The CDN wake word detector analyzes the audio stream to determine if the audio stream contains any audio that sounds like the wake word. If so, the CDN wake word detector can generate metadata that describes the time period, within the audio content, in which the false wake word was encountered. The metadata can include time offsets, from the start of the audio content, which can instruct a voice-enabled device to deactivate during the time period. This metadata is stored and then sent to the media-playback device requests the media content. The media-playback device can then instruct or inform the voice-enabled device of the presence of the false wake word. In this way, the wake word detector, at the voice-enabled device, is not activated to receive the false wake word.