Temporal decorrelation method for robust speaker verification

摘要

A system and method for automated temporal decorrelation of speaker voice verification is shown including a collector (210) for receiving speech inputs from an unknown speaker claiming a specific identity, a word-level speech feature calculator (220) for generating word-level speech feature vectors from such speech inputs, word-level speech feature storage (170) for storing word-level speech feature vectors known to belong to a speaker with the specific identity, a word-level vector scorer (230) to calculate a similarity score between the word-level speech feature vectors received from the unknown speaker with those received from the word-level speech feature storage, and speaker verification decision circuitry (240) for determining, based on the similarity score, whether the unknown speaker's identity is the same as that claimed. The word-level vector scorer further includes concatenation circuitry as well as a word-specific orthogonalizing linear transformer (160-Fig. 4b; 380-Fig. 7).;Other systems and methods are also disclosed.