Inference-Aided Speaker Recognition Using Contextual Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud-based services, accurately identifying the speaker's voice is challenging, especially when the communication device is owned or controlled by someone other than the intended user, leading to potential misidentification and incorrect access to personal information.
Innovation Solution
Implementing inference-aided speaker recognition that compares the received voice utterance against a limited group of potential speakers identified through connections such as location and social connections, reducing computational intensity by limiting the universe of potential speakers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system compares voice information against all users in the cloud-based service, then the speaker recognition accuracy is improved, but the computational burden on servers increases significantly
Solution Approach 1:
The patent segments the set of all users into a smaller comparison group by first filtering users based on contextual information such as device ownership, location, and social connections. This segmentation reduces the number of voice signatures that need to be compared against each new voice input, thereby reducing server computational burden while maintaining acceptable recognition accuracy within the narrowed candidate set.
Solution Approach 2:
The patent applies local quality by tailoring the comparison group to the specific context of each voice input. Instead of using a uniform comparison approach for all users, the system adjusts the comparison group based on local contextual factors such as which device the voice is associated with, the user's location, and their social connections, thereby optimizing the balance between accuracy and computational efficiency for each specific case.
2Productivity
If the system uses contextual information to limit the comparison group, then the computational burden is reduced, but the risk of misidentification increases
Solution Approach 1:
The patent incorporates feedback mechanisms where the system continuously refines the comparison group based on the results of voice comparisons and user feedback. When a voice match is found within the limited comparison group, the system can verify the identification through additional contextual checks or user confirmation, thereby maintaining high reliability while benefiting from the reduced computational scope of the limited comparison group.
Data Source
AI summary
Systems, methods performed by data processing apparatus and computer storage media encoded with computer programs for receiving information relating to (i) a communication device that has received an utterance and (ii) a voice associated with the received utterance, comparing the received voice information with voice signatures in a comparison group, the comparison group including one or more individuals identified from one or more connections arising from the received information relating to the communication device, attempting to identify the voice associated with the utterance as matching one of the individuals in the comparison group, and based on a result of the attempt to identify, selectively providing the communication device with access to one or more resources associated with the matched individual.