Speech User Recognition via ASR Confidence and Location Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional speech recognition systems face challenges in accurately verifying the identity of users, leading to potential misidentification and unauthorized access, especially when multiple users speak the same utterance with different intentions.
Innovation Solution
The system incorporates a confidence output from the Automatic Speech Recognition (ASR) component and secondary information such as device location and image data into the user recognition analysis, enhancing the accuracy of user verification by comparing speech characteristics with stored training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional speech recognition systems are used for user verification, then the system is simple and easy to operate, but the accuracy of user verification is low leading to false verifications
Solution Approach 1:
The patent combines multiple verification components into a unified system: ASR confidence output, speaker verification module, and secondary information processing are merged to work together. The speaker verification module integrates audio feature extraction, template matching, and confidence threshold evaluation to achieve accurate user verification while managing system complexity through modular design.
2Reliability
If only speech characteristics are used for user recognition, then the system is simple, but it cannot distinguish between multiple users speaking the same utterance with different intentions
Solution Approach 1:
The patent segments the verification process into distinct components: ASR processing for utterance recognition, speaker verification for identity confirmation, and secondary information analysis for contextual understanding. Each component processes specific features independently before integrating results, allowing the system to handle complex verification scenarios without overwhelming complexity in any single module.
Solution Approach 2:
The patent adds multiple verification dimensions beyond simple speech recognition. It incorporates speaker identity verification, confidence level assessment, secondary information analysis (device location, usage patterns), and intent determination. This multi-dimensional approach enables the system to distinguish between multiple users speaking the same utterance by evaluating different aspects simultaneously.
3Reliability
If speech recognition systems process all utterances, then productivity is high, but false verifications lead to unintended data disclosure
Solution Approach 1:
The patent performs preliminary verification actions before processing utterances. The speaker verification module checks audio characteristics against stored templates in advance, and the confidence threshold filter evaluates ASR confidence levels before triggering full processing. This preliminary screening prevents false verifications from reaching subsequent processing stages, reducing unintended data disclosure while maintaining high productivity by filtering out low-confidence cases early.
Data Source
AI summary
Systems, methods, and devices for verifying a user are disclosed. A speech-controlled device captures a spoken command, and sends audio data corresponding thereto to a server. The server performs ASR on the audio data to determine ASR confidence data. The server, in parallel, performs user verification on the audio data to determine user verification confidence data. The server may modify the user verification confidence data using the ASR confidence data. In addition or alternatively, the server may modify the user verification confidence data using at least one of a location of the speech-controlled device within a building, a type of the speech-controlled device, or a geographic location of the speech-controlled device.