Dynamic Speech Grammar Distractor Selection via Acoustic Dissimilarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition grammars with static distractors are ineffective in accurately rejecting incorrect utterances due to varying degrees of dissimilarity, leading to potential false acceptance or rejection in identity verification processes.
Innovation Solution
A system dynamically generates speech recognition grammars by selecting distractors based on acoustic characteristics of a target entry, enhancing the likelihood of correctly rejecting non-matching utterances through a dynamic grammar builder that includes an analyzing module for acoustic dissimilarity analysis and a grammar-generating module for selecting appropriate distractors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static distractors are used in speech recognition grammars, then the grammar structure is simple and easy to implement, but the accuracy of rejecting incorrect utterances deteriorates due to varying degrees of dissimilarity
Solution Approach 1:
The patent applies dynamics by transitioning from static distractors to dynamic distractor selection. The system dynamically selects distractors based on acoustic characteristics of the target entry and real-time analysis of speech signals. The grammar generator creates multiple possible grammars with different distractor combinations and selects the optimal one based on acoustic dissimilarity measures, enabling adaptive rejection of incorrect utterances while maintaining manageable complexity through algorithmic automation.
Solution Approach 2:
The patent changes parameters by using acoustic dissimilarity measures as selection criteria for distractors. Instead of using fixed distractors, the system calculates acoustic dissimilarity between potential distractors and the target entry, then selects distractors based on this parameter. This parameter-based approach allows the system to adapt to different speech patterns and acoustic conditions, improving rejection accuracy without requiring complex manual grammar design.
2Reliability
If conventional static distractors are used, then the system is simple to operate, but false acceptance or rejection occurs due to insufficient acoustic dissimilarity analysis
Solution Approach 1:
The system applies self-service by automatically generating and selecting optimal grammars with appropriate distractors. The grammar generator autonomously analyzes acoustic characteristics, calculates dissimilarity measures, and selects the best distractor combinations without requiring manual intervention. This automation maintains ease of operation while significantly reducing false acceptances and rejections through data-driven, algorithmic decision-making processes.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously analyzes acoustic characteristics of speech signals and adjusts distractor selection accordingly. The analysis module provides feedback on acoustic dissimilarity measures, which the grammar generator uses to refine distractor selection. This feedback loop enables the system to adapt to varying acoustic conditions and maintain high reliability in rejecting incorrect utterances while operating simply through automated processes.
3Productivity
If a single static set of distractors is used, then the grammar is simple and fast to process, but the effectiveness of distractors varies depending on acoustic dissimilarity
Solution Approach 1:
The system dynamically adapts distractor selection based on acoustic characteristics rather than using a fixed set. The grammar generator creates multiple potential grammars with different distractor combinations and selects the optimal one based on real-time acoustic analysis. This dynamic approach maintains processing efficiency by automating the selection process while ensuring consistent distractor effectiveness across different acoustic conditions and speech patterns.
Solution Approach 2:
The patent changes the parameter of distractor selection from static to dynamic based on acoustic dissimilarity measures. The system calculates acoustic dissimilarity between potential distractors and the target entry, then selects distractors that optimize recognition performance. This parameter-based dynamic selection ensures consistent effectiveness across varying acoustic conditions while maintaining productivity through algorithmic automation that processes selections efficiently.
Data Source
AI summary
A computer-implemented method for dynamically generating a speech recognition grammar is provided. The method includes determining a target entry, and accessing a plurality of potential distracters. The method also includes selecting one or more distracters from the plurality of potential distracters. More particularly, each potential distracter selected is selected based upon an assessed acoustic dissimilarity between the distracter and the target entry. The method further includes dynamically generating a speech recognition grammar that includes the target entry and one or more of the distracters selected based upon an acoustic dissimilarity to the target entry.