Evaluating speaker recognition performance

• Main Authors: PELECANOS JASON, HUANG YILING, WANG QUAN, CHAO PU-SEN
• Format: Patent
• Language: Chinese; English
• Published: 18.07.2023
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

A method (300) for evaluating a verification model (146) includes receiving a first set of verification results and a second set of verification results (148), wherein each verification result indicates whether a master model or an alternative model verifies an identity of a user (10) as a registered user. The method further includes identifying the first set of verification results and each verification result in the second set including a performance metric (212); determining a first score (222) for the master model based on a plurality of verification results identified in the first set including performance metrics; and determining a second score for the alternative model based on the plurality of verification results identified in the second set including performance metrics. The method further includes determining whether a verification capability of the alternative model is superior to a verification capability of the master model based on the first score and the second score. 一种用于评估验证模型(146)的方法(300)包括