Compensation for phonetic nuisance variability in speaker recognition using DNNs

In this paper, a new way of using phonetic DNN in textindependent speaker recognition is examined. Inspired by the Subspace GMM approach to speech recognition, we try to extract i-vectors that are invariant to the phonetic content of the utterance. We overcome the assumption of Gaussian distributed senones by combining DNN with UBM posteriors and we form a complete EM algorithm for training and extracting phonetic content compensated i-vectors. A simplified version of the model is also presented, where the phonetic content and speaker subspaces are learned in a decoupled way. Covariance adaptation is also examined, where the covariance matrices are reestimated rather than copied from the UBM. A set of primary experimental results is reported on NIST-SRE 2010, with modest improvement when fused with the standard i-vectors.