In this study, we have proposed and tested a new extension of the PLDA model, where within-class (channel) variability is modeled as a function of the class (speaker) location in the feature (iVector) space.
In this study, we have proposed and tested a new extension of the PLDA model, where within-class (channel) variability is modeled as a function of the class (speaker) location in the feature (iVector) space. (en)