Duration dependent covariance regularization in PLDA modeling for speaker verification

In this paper, we present a covariance regularized probabilistic linear discriminant analysis (CR-PLDA) model for text independent speaker verification. In the conventional simplified PLDA modeling, the covariance matrix used to capture the residual energies is globally shared for all i-vectors. However, we believe that the point estimated i-vectors from longer speech utterances may be more accurate and their corresponding covariances in the PLDA modeling should be smaller. Similar to the inverse 0 order statistics weighted covariance in the i-vector model training, we propose a duration dependent normalized exponential term containing the duration normalizing factor μ and duration extent factor ν to regularize the covariance in the PLDAmodeling. Experimental results are reported on the NIST SRE 2010 common condition 5 female part task and the NIST 2014 i-vector machine learning challenge, respectively. For both tasks, the proposed covariance regularized PLDA system outperforms the baseline PLDA system by more than 13% relatively in terms of equal error rate (EER) and norm minDCF values.