Speaker Recognition Using Segmented Phrase Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speaker recognition methods face high error rates when using unique passphrases and require separate operations for identity verification, which are inefficient and less accurate compared to common passphrase verification.
Innovation Solution
The method involves prompting users to speak a phrase containing both a personal identifier and a common phrase component, using keyword spotting to extract the personal identifier, and comparing the common phrase component against stored voice prints to verify the user, thereby achieving higher accuracy without the need for a separate identity claim operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If unique passphrases are used for speaker recognition, then identity verification accuracy is improved, but error rates increase and separate operations are required
Solution Approach 1:
The spoken phrase is segmented into two distinct components: a personal identifier portion and a common phrase portion. The personal identifier (e.g., name, account number) is extracted and verified against stored identifying information, while the common phrase portion is used for voice print comparison. This segmentation allows the system to leverage the accuracy of unique identifiers while maintaining the reliability of common phrase verification.
2Measurement precision
If unique passphrases are used for speaker recognition, then identity verification accuracy is improved, but system complexity increases due to separate operations
Solution Approach 1:
The system merges the identity verification function and the speaker verification function into a single integrated operation. By combining the personal identifier extraction with the common phrase voice print comparison, the system performs both identification and verification simultaneously, eliminating the need for separate operations and reducing overall system complexity.
Solution Approach 2:
The spoken phrase serves multiple functions simultaneously: it provides the personal identifier for identity lookup and contains the common phrase component for voice biometric verification. This multi-functionality allows a single user input to trigger both identification and verification processes, streamlining the overall system operation.
3Productivity
If common passphrases are used for speaker recognition, then operation efficiency is improved, but verification accuracy decreases
Solution Approach 1:
The phrase is segmented into personal identifier and common phrase components, allowing the system to use the common phrase for efficient voice print matching while the personal identifier provides the necessary verification accuracy through comparison against stored identifying information.
Data Source
AI summary
Embodiments of the present invention perform speaker identification and verification by first prompting a user to speak a phrase that includes a common phrase component and a personal identifier. Then, the embodiments decompose the spoken phrase to locate the personal identifier. Finally, the embodiments identify and verify the user based on the results of the decomposing.