Voice Biometric Authentication Using Context-Limited Identity Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice biometrics systems require separate enrollment with each vendor system, limiting user experience and adoption across diverse devices and public settings, and face challenges in resource-intensive identification processes and false acceptance/rejection rates.
Innovation Solution
A centralized machine-learning architecture enables seamless user authentication across multiple devices and services by generating biometric and context embeddings, allowing enrollment once and authenticating across disparate systems, while mitigating resource demands and false acceptance/rejection rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voice biometrics systems are deployed privately with single-vendor enrollment, then authentication accuracy is improved, but user experience and adoption across multiple vendors deteriorates
Solution Approach 1:
The patent introduces a centralized authentication server as an intermediary between voice devices and users. This server maintains a centralized database of voice prints and handles authentication requests from multiple vendors uniformly. The intermediary architecture allows each vendor to maintain their own authentication system while the centralized server provides cross-vendor interoperability, resolving the contradiction between authentication accuracy and cross-vendor compatibility.
Solution Approach 2:
The centralized authentication server implements universality by serving multiple vendors and device types through a single unified system. The server can authenticate users across different voice devices (smart speakers, telephones, video conferencing systems) without requiring separate enrollment processes for each vendor, thereby achieving both high authentication accuracy and broad adaptability.
2Adaptability or versatility
If centralized authentication system is implemented, then cross-device authentication is enabled, but system complexity increases
Solution Approach 1:
The system is segmented into distinct functional components: voice devices that capture audio, a centralized authentication server that processes authentication requests, and a database that stores voice prints. This segmentation allows each component to be optimized independently and simplifies the overall architecture by distributing complexity across manageable modules rather than requiring monolithic integration.
Solution Approach 2:
The centralized authentication server acts as an intermediary that abstracts the complexity of cross-device authentication from individual vendors and devices. By centralizing the authentication logic in a dedicated server, the patent reduces the complexity burden on each individual device while enabling seamless cross-device authentication functionality.
3Reliability
If comprehensive voice biometric enrollment is performed, then authentication security is improved, but resource consumption and processing time increases
Solution Approach 1:
The system performs preliminary action by capturing and storing voice prints during the enrollment phase before actual authentication is needed. During authentication, the system only needs to compare the incoming voice print against the stored enrollment data, significantly reducing the computational resources required compared to performing comprehensive voice analysis in real-time during authentication.
Solution Approach 2:
The patent creates a copy of the voice print during enrollment and stores it in the database. During authentication, the system compares the incoming voice sample against this pre-created copy rather than performing comprehensive re-analysis. This copying approach maintains high authentication security through accurate voice print comparison while dramatically reducing processing resource consumption during actual authentication events.
Data Source
AI summary
Disclosed are systems and methods including computing-processes executing machine-learning architectures extract vectors representing disparate types of data and output predicted identities of users accessing computing services, without express identity assertions, and across multiple computing services, analyzing data from multiple modalities, for various user devices, and agnostic to architectures hosting the disparate computing service. The system invokes the identification operations of the machine-learning architecture, which extracts biometric embeddings from biometric data and context embeddings representing all or most of the types of metadata features analyzed by the system. The context embeddings help identify a subset of potentially matching identities of possible users, which limits the number of biometric-prints the system compares against an inbound biometric embedding for authentication. The types of extracted features originate from multiple modalities, including metadata from data communications, audio signals, and images. In this way, the embodiments apply a multi-modality machine-learning architecture.