Two-Tier Voice Biometric Speaker Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems face challenges in accurately identifying enrolled speakers, particularly in distinguishing between enrolled and non-enrolled users, with existing voice biometric processes often balancing between False Acceptance Rate, False Rejection Rate, and Equal Error Rate, and requiring efficient power management.
Innovation Solution
A method and system that employ a two-tiered voice biometric process, where a first voice biometric process with lower power consumption and computational complexity is used for initial identification, followed by a more discriminative second process if the initial determination suggests a potential match, utilizing various techniques such as neural nets, Gaussian Mixture Models, and i-vector processes to enhance accuracy and reduce errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single high-accuracy voice biometric process is used, then speaker identification accuracy is improved, but power consumption and computational complexity increase
Solution Approach 1:
The voice biometric verification process is segmented into two distinct stages: a first voice biometric process for initial verification and a second voice biometric process for final verification. This segmentation allows the system to use a computationally lighter first process to filter out obvious non-matches, thereby reducing the overall power consumption compared to running only the more intensive second process on all inputs.
Solution Approach 2:
The system dynamically selects which voice biometric process to execute based on the output of the first process. If the first process indicates a potential match, the system activates the more computationally intensive second process. This dynamic adaptation allows the system to optimize power consumption by running the heavy second process only when necessary, rather than continuously.
2Measurement precision
If a single high-accuracy voice biometric process is used, then speaker identification accuracy is improved, but device complexity increases
Solution Approach 1:
The verification system is divided into two sequential processes with distinct functions. The first process handles initial filtering and basic verification, while the second process handles final confirmation. This segmentation reduces device complexity by organizing the system into manageable modules with clear responsibilities, making the overall system easier to implement and maintain compared to a single monolithic high-accuracy process.
Solution Approach 2:
The first voice biometric process acts as an intermediary between the audio input and the second voice biometric process. It prepares and pre-processes the verification request, filtering out obvious non-matches before passing candidates to the second process. This intermediary role simplifies the architecture by creating a clear pipeline structure and reducing the computational burden on the final verification stage.
3Use of energy by moving object
If a two-tiered voice biometric process is used, then power efficiency is improved, but processing time increases
Solution Approach 1:
The first voice biometric process performs preliminary verification before the second process is activated. This preliminary action quickly eliminates obvious non-matches, so that the more time-consuming second process is only executed when there is a genuine possibility of a match. This preliminary filtering reduces the average processing time across all verification attempts, even though the two-process system is more complex than a single-process system.
4Reliability
If a more discriminative second voice biometric process is used, then False Acceptance Rate is reduced, but computational complexity increases
Solution Approach 1:
The verification process is segmented into two stages with different accuracy requirements. The first process provides initial verification with lower computational complexity, while the second process provides final verification with higher discriminative power. This segmentation allows the system to achieve low False Acceptance Rates through the combination of both processes, rather than requiring a single extremely complex process to handle all verification needs.
Solution Approach 2:
The system dynamically activates the more computationally intensive second voice biometric process only when the first process indicates a potential match. This dynamic activation ensures that the high-complexity second process is run selectively rather than continuously, reducing the average computational complexity while maintaining the reliability benefits of having a highly discriminative verification stage.
Data Source
AI summary
A method of speaker identification comprises receiving an audio signal representing speech; performing a first voice biometric process on the audio signal to attempt to identify whether the speech is the speech of an enrolled speaker; and, if the first voice biometric process makes an initial determination that the speech is the speech of an enrolled user, performing a second voice biometric process on the audio signal to attempt to identify whether the speech is the speech of the enrolled speaker. The second voice biometric process is selected to be more discriminative than the first voice biometric process.