INTERSPEECH 2013

Total variability modeling, based on ivector extraction of converting a variablelength sequence of feature vectors into a fixedlength ivector, is currently an adopted parametrization technique for state oftheart speaker verification systems. However, when the number of the feature vectors is low, uncertainty in the ivector representation as a point estimate of the linearGaussian model is understandably problematic. It is known that the zeroth and first order sufficient statistics, given the hyperparameters, completely characterize the extracted ivectors. In this study we propose to use a minimax strategy to estimate the sufficient statistics in order to increase the robustness of the extracted ivectors. We show by experiments that the proposed minimax technique can improve over the baseline system from 9.89% to 7.99% on the NIST SRE 2010 8conv10sec task.
Bibliographic reference. Hautamäki, Ville / Cheng, YouChi / Rajan, Padmanabhan / Lee, ChinHui (2013): "Minimax ivector extractor for short duration speaker verification", In INTERSPEECH2013, 37083712.