Local Speaker Recognition via Voice Print Pruning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speaker recognition technologies are resource-intensive and unsuitable for mobile devices, requiring remote server processing and transmission of voice data, which leads to bandwidth issues and security risks, and lack user-friendly and cost-effective voice biometric solutions for secure authentication.
Innovation Solution
A mobile application that enables users to create and refine voice prints on their devices, allowing local voice authentication with reduced resource demands by pruning statistical data and using a universal background model to generate and compare voice biometric prints, enabling secure and efficient speaker recognition without centralized storage of voice biometrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speaker recognition is performed using remote server processing, then authentication accuracy can be maintained, but bandwidth consumption increases and security risks arise from centralized storage
Solution Approach 1:
The patent extracts the voice biometric processing functionality from the remote server and implements it locally on mobile devices. The system prunes statistical data and uses a universal background model to generate voice biometric prints locally, eliminating the need to transmit voice data to remote servers while maintaining authentication accuracy
Solution Approach 2:
The patent changes the parameter of data representation by pruning statistical data and using compact universal background models. This allows the system to perform speaker recognition with reduced data requirements, enabling local processing without consuming significant bandwidth
2Ease of operation
If voice biometric data is stored centrally on remote servers, then authentication can be performed, but security risks and vulnerability to data breaches increase
Solution Approach 1:
The patent extracts voice biometric data storage from centralized remote servers and places it locally on individual mobile devices. Each device stores its own pruned statistical data and universal background model, eliminating the security risks associated with centralized storage while maintaining full authentication functionality
Solution Approach 2:
The patent segments the centralized voice biometric storage system into distributed local storage units on individual mobile devices. Each device maintains its own independent voice print data, creating a distributed architecture that improves security while preserving authentication capabilities
3Measurement precision
If full statistical data is used for speaker recognition, then recognition accuracy is maintained, but resource consumption on mobile devices becomes too high
Solution Approach 1:
The patent extracts and removes unnecessary statistical data from the complete dataset, retaining only the essential features needed for speaker recognition. This pruning process reduces resource consumption while preserving recognition accuracy by keeping only the most relevant statistical parameters
Solution Approach 2:
The patent applies local quality optimization by selectively retaining specific statistical data that is most valuable for speaker recognition while discarding less important data. The universal background model is configured with optimized parameters that provide sufficient accuracy for mobile device processing
Data Source
AI summary
A speaker recognition system for authenticating a mobile device user includes an enrollment and learning software module, a voice biometric authentication software module, and a secure software application. Upon request by a user of the mobile device, the enrollment and learning software module displays text prompts to the user, receives speech utterances from the user, and produces a voice biometric print. The enrollment and training software module determines when a voice biometric print has met at least a quality threshold before storing it on the mobile device. The secure software application prompts a user requiring authentication to repeat an utterance based at least on an attribute of a selected voice biometric print, receives a corresponding utterance, requests the voice biometric authentication software module to verify the identity of the second user using the utterance, and, if the user is authenticated, imports the voice biometric print.