Wake Word Recognition via Phonetic Similarity Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital assistants require precise pronunciation of wake words or phrases for activation, which can be challenging for individuals with phonological limitations, leading to user frustration and delays.
Innovation Solution
The method involves receiving user input as a potential wake word, determining its association with a stored wake word through phonetic variations, and activating the digital assistant accordingly, allowing for activation even if the user cannot pronounce the exact wake word.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional digital assistants require precise pronunciation of wake words for activation, then activation reliability is improved, but ease of operation deteriorates for users with phonological limitations
Solution Approach 1:
The system changes the parameter of wake word recognition from exact string matching to phonetic similarity matching. By converting wake words and user inputs into phonetic representations and comparing similarity scores, the system accepts variations in pronunciation while maintaining reliable activation. This resolves the contradiction by allowing users with phonological limitations to activate the assistant easily without compromising activation reliability.
Solution Approach 2:
The patent introduces phonetic transcription as an intermediary between the user's spoken input and the wake word recognition system. Instead of directly comparing raw audio or text strings, the system converts both to phonetic representations and uses phonetic similarity as a mediator to determine if the input matches the wake word. This intermediary layer enables flexible recognition that accommodates pronunciation variations while maintaining systematic control over activation.
2Ease of operation
If the system accepts phonetic variations of wake words, then ease of operation is improved for users with phonological limitations, but measurement precision deteriorates in wake word recognition
Solution Approach 1:
The system changes the recognition parameter from exact string matching to phonetic similarity scoring. By representing wake words and user inputs in phonetic form and calculating similarity metrics, the system can tolerate pronunciation variations while maintaining controlled precision through configurable similarity thresholds. This allows the system to balance accessibility for users with phonological limitations against accurate wake word recognition.
Solution Approach 2:
The wake word recognition system becomes dynamic by adjusting its acceptance criteria based on phonetic similarity scores rather than fixed exact matching rules. The system can adaptively determine whether an input qualifies as a wake word activation by evaluating phonetic proximity, allowing flexibility in pronunciation while maintaining systematic control over when activation occurs. This dynamic approach resolves the tension between accepting variations and maintaining precision.
3Measurement precision
If conventional systems use exact wake word matching, then wake word recognition precision is improved, but loss of time increases due to user frustration and repeated attempts
Solution Approach 1:
The system changes from exact string matching to phonetic similarity comparison, which processes inputs more efficiently by working with compressed phonetic representations rather than full audio or text analysis. This approach maintains controlled recognition precision through similarity thresholds while reducing the time users spend on repeated activation attempts, thereby decreasing overall activation time and eliminating frustration associated with exact pronunciation requirements.
Data Source
AI summary
One embodiment provides a method, including: receiving, at an information handling device, user input comprising a potential wake word; determining, using a processor, whether the potential wake word is associated with a stored wake word; and responsive to determining that the potential wake word is associated with the stored wake word, activating, based on the potential wake word, a digital assistant associated with the information handling device. Other aspects are described and claimed.