Voice Authentication Biasing Parameters for Assistant Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user authentication techniques for automated assistants, such as text-dependent and text-independent speaker authentication, face challenges in accurately verifying users, especially in multi-user environments with similar voice characteristics, leading to prolonged interactions and resource utilization.
Innovation Solution
Implementing biasing parameters on a per-user and device-by-device basis, taking into account past interactions and contextual conditions, to enhance the accuracy and robustness of speaker authentication, thereby reducing the need for additional authentication methods and preventing errant verifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If text-dependent or text-independent speaker authentication is used to verify users, then user authentication capability is provided, but authentication accuracy deteriorates in multi-user environments with similar voice characteristics
Solution Approach 1:
The patent modifies the authentication decision parameters by introducing biasing parameters that adjust the threshold for verifying user identity. Instead of using a fixed distance metric threshold, the system dynamically adjusts the authentication criteria based on device-specific and user-specific factors, allowing more accurate differentiation between users with similar voice characteristics.
Solution Approach 2:
The patent applies different authentication criteria and biasing parameters to different devices and users individually. Each device-microphone combination and user pair receives customized authentication parameters based on their specific characteristics, rather than applying a uniform authentication standard across all users and devices.
2Ease of operation
If traditional speaker authentication is used, then authentication process is simple, but interaction time increases due to repeated authentication failures
Solution Approach 1:
The patent performs preliminary characterization of each device-microphone combination and user pair during enrollment, storing biasing parameters that will be applied during subsequent authentication. This preliminary action prepares the authentication system in advance, preventing the need for repeated authentication attempts and reducing interaction time.
Solution Approach 2:
The system uses feedback from past authentication interactions and device characteristics to dynamically adjust biasing parameters. By analyzing authentication outcomes and device performance, the system refines its authentication criteria over time, reducing failures and minimizing the need for repeated authentication attempts.
3Reliability
If multiple authentication attempts are made for verification, then authentication thoroughness increases, but resource utilization increases
Solution Approach 1:
The patent changes the authentication parameter from a uniform threshold to dynamic biasing parameters that account for device and user characteristics. This allows the system to achieve thorough verification with fewer authentication attempts by using more accurate, customized thresholds from the beginning, thereby reducing resource utilization.
4Measurement precision
If device-specific authentication parameters are used, then authentication accuracy for each device improves, but system complexity increases
Solution Approach 1:
The patent implements device-specific authentication by applying local biasing parameters to each device-microphone combination and user pair. Rather than creating complex device-specific authentication systems, the approach uses simple parameter adjustments that are applied uniformly across the authentication logic, maintaining system simplicity while improving per-device accuracy.
Data Source
AI summary
Implementations are directed to biasing speaker authentication on a per-user basis and on a device-by-device basis and/or contextual feature(s) basis. In some of those implementations, in performing speaker authentication based on a spoken utterance, different biasing parameters are determined for each of multiple different registered users of an assistant device at which the spoken utterance was detected. In those implementations, each of the biasing parameters can be used to make it more likely or less likely (in dependence of the biasing parameter) that a corresponding registered user will be verified using the speaker authentication. Through utilization of biasing parameter(s) in performing speaker authentication, accuracy and/or robustness of speaker authentication can be increased.