Dynamic Wake Word Detection Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice-enabled smart devices face challenges in accurately determining wake words and ensuring that recognized speech originates from authorized users, leading to potential misinterpretation and unauthorized access.
Innovation Solution
A system and method where a user device detects speech, assigns a confidence score to potential wake words, and adjusts a wake word detection threshold based on whether the speech is from an authorized user by generating and comparing voiceprints, thereby increasing or decreasing scrutiny for wake word recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed wake word detection threshold is used, then the device can operate with simple detection logic, but it cannot adapt to different users and may produce false positives or false negatives
Solution Approach 1:
The wake word detection threshold is transformed from a fixed value to a dynamic value that changes based on user identity. The system generates a voiceprint from detected speech, compares it to stored voiceprints to identify the user, and then selects or adjusts the detection threshold accordingly. This allows the threshold to adapt to different users' speaking characteristics while maintaining simple detection logic for each user.
Solution Approach 2:
The detection threshold parameter is changed based on the identified user's voice characteristics. Different users have different authorized threshold values stored in the system. When a user is identified through voiceprint matching, the system retrieves and applies that user's specific threshold parameter, enabling adaptability without requiring complex real-time adjustments.
2Measurement precision
If the wake word detection threshold is lowered to increase sensitivity, then more speech is recognized as wake words, but false positives from unauthorized users or background noise increase
Solution Approach 1:
The system performs preliminary user identification through voiceprint analysis before final wake word recognition. By first generating a voiceprint from the detected speech and comparing it against stored voiceprints to identify the user, the system can then apply an appropriate detection threshold. This preliminary action ensures that even if the threshold is low, only authorized users can trigger wake word recognition, preventing false positives from unauthorized sources.
Solution Approach 2:
Voiceprint matching serves as an intermediary verification step between the detected speech and the wake word recognition decision. The system uses the voiceprint comparison result to determine user authorization status, which then influences the wake word detection threshold application. This intermediary mechanism separates the sensitivity of wake word detection from the security of user authorization.
3Reliability
If the wake word detection threshold is raised to reduce false positives, then unauthorized access is prevented, but legitimate wake words from authorized users may be missed
Solution Approach 1:
The system performs preliminary user identification through voiceprint analysis before applying the wake word detection threshold. By first determining whether the speech originates from an authorized user through voiceprint matching, the system can then apply the appropriate threshold level. This ensures that authorized users benefit from lower, more sensitive thresholds while unauthorized users face higher, more restrictive thresholds.
Solution Approach 2:
Different detection threshold values are applied to different users based on their individual voice characteristics and authorization status. Each authorized user has their own optimized threshold stored in the system. The system applies the appropriate local threshold to each user's speech independently, allowing high sensitivity for authorized users while maintaining security against unauthorized access.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A user device (e.g., voice assistant device, voice enabled device, smart device, computing device, etc.) may receive/detect audio content (e.g., speech, etc.) that includes a wake word and/or words similar to a wake word. The user device may require a wake word, a portion of the wake word, or words similar to the wake word to be detected prior to interacting with a user. The user device may, based on characteristics of the audio content, determine if the audio content originates from an authorized user. The user device may decrease and/or increase scrutiny applied to wake word detection based on whether audio content originates from an authorized user.