Speaker Verification Using Dynamic Vocabulary Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speaker verification systems are vulnerable to playback attacks and suffer from accuracy issues due to voice changes over time, illness, and environmental factors, as they often use the same vocabulary for enrollment and challenge utterances, making it difficult to distinguish between genuine and fraudulent access attempts.
Innovation Solution
Implementing a high-perplexity vocabulary that differs from the enrollment vocabulary, combined with adaptation techniques within a single verification session to update voice prints and account for voice changes, and incorporating speech recognition to verify the accuracy of challenge utterances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the same vocabulary is used for enrollment and challenge utterances, then verification accuracy is improved, but the system becomes vulnerable to playback attacks
Solution Approach 1:
The system dynamically changes the vocabulary used for challenge utterances based on the enrollment phase. During enrollment, a first vocabulary is used to build the voice print, while during verification, a different second vocabulary is used for challenge utterances. This dynamic switching prevents playback attacks while maintaining verification accuracy through adaptive vocabulary selection.
Solution Approach 2:
The system changes the lexical parameter (vocabulary) between enrollment and verification phases. By transforming the vocabulary set used for challenge utterances to differ from the enrollment vocabulary, the system alters the linguistic parameters while preserving the acoustic characteristics needed for accurate speaker verification.
2Reliability
If a different vocabulary is used for challenge utterances, then security against playback attacks is improved, but verification accuracy deteriorates
Solution Approach 1:
The system uses feedback from the voice signal analysis to adapt the voice print during the verification session. By continuously updating the voice model based on the challenge utterance responses, the system compensates for the vocabulary change and maintains accurate speaker verification despite using a different vocabulary set.
Solution Approach 2:
The system performs preliminary adaptation of the voice print using the challenge utterance before final verification. This preliminary action allows the system to adjust to the different vocabulary and acoustic conditions, ensuring accurate verification while maintaining security against playback attacks.
3Measurement precision
If voice print adaptation is performed within a single session, then accuracy is improved by accounting for voice changes, but system complexity increases
Solution Approach 1:
The system merges the enrollment and verification phases into a single integrated session where voice print adaptation occurs continuously. By combining these functions and performing real-time adaptation during verification, the system improves accuracy for voice changes while managing complexity through unified processing rather than separate systems.
Data Source
AI summary
In one aspect, a method for determining a validity of an identity asserted by a speaker using a voice print is provided. The method comprises acts of performing a first verification stage comprising comparing a first voice signal from the speaker uttering at least one first challenge utterance with at least a portion of the voice print and performing a second verification stage if it is concluded in the first verification stage that the first voice signal was obtained from an utterance by the user. The second verification stage comprises adapting at least one parameter of the voice print based, at least in part, on the first voice signal to obtain an adapted voice print, and comparing a second voice signal from the speaker uttering at least one second challenge utterance with at least a portion of the adapted voice print.