Speakers In The Wild Research Articles

Coupling a Generative Model With a Discriminative Learning Framework for Speaker Verification

The task of speaker verification (SV) is to decide whether an utterance is spoken by a target or an imposter speaker. In most studies of SV, a log-likelihood ratio (LLR) score is estimated based on a generative probability model on speaker features, and compared with a threshold for making a decision. However, the generative model usually focuses on individual feature distributions, does not have the discriminative feature selection ability, and is easy to be distracted by nuisance features. The SV, as a hypothesis test, could be formulated as a binary discrimination task where neural network based discriminative learning could be applied. In discriminative learning, the nuisance features could be removed with the help of label supervision. However, discriminative learning pays more attention to classification boundaries, and is prone to overfitting to a training set which may result in bad generalization on a test set. In this paper, we propose a hybrid learning framework, i.e., coupling a joint Bayesian (JB) generative model structure and parameters with a neural discriminative learning framework for SV. In the hybrid framework, a two-branch Siamese neural network is built with dense layers that are coupled with factorized affine transforms as used in the JB model. The LLR score estimation in the JB model is formulated according to the distance metric in the discriminative learning framework. By initializing the two-branch neural network with the generatively learned model parameters of the JB model, we further train the model parameters with the pairwise samples as a binary discrimination task. Moreover, a direct evaluation metric (DEM) in SV based on minimum empirical Bayes risk (EBR) is designed and integrated as an objective function in the discriminative learning. We carried out SV experiments on Speakers in the wild (SITW) and Voxceleb. Experimental results showed that our proposed model improved the performance with a large margin compared with state of the art models for SV.

13 years of speaker recognition research at BUT, with longitudinal analysis of NIST SRE

In this paper, we present a brief history and a “longitudinal study” of all important milestone modelling techniques used in text independent speaker recognition since Brno University of Technology (BUT) first participated in the NIST Speaker Recognition Evaluation (SRE) in 2006—GMM MAP, GMM MAP with eigen-channel adaptation, Joint Factor Analysis, i-vector and DNN embedding (x-vector). To emphasize the historical context, the techniques are evaluated on all NIST SRE sets since 2004 on a time-machine principle, i.e. a system is always trained using all data available up till the year of evaluation. Moreover, as user-contributed audiovisual content dominates nowadays’ Internet, we representatively include the Speakers In The Wild (SITW) and VOiCES challenge datasets in the evaluation of our systems. Not only we present a comparison of the modelling techniques, but we also show the effect of sampling frequency.

Total Variability Layer in Deep Neural Network Embeddings for Speaker Verification

The total variability model (TVM) has been extensively used as a tool to obtain a vector representation of the sources of variability present in a signal. However, recent studies have shown that embeddings derived from a deep neural network (DNN) architecture can provide significant performance improvement over TVM for the speaker verification task. In this letter, we show that TVM can also be reformulated in a manner that enables the integration of a DNN within the model. In addition, we show that this TVM architecture can also be incorporated as one of the layers within a DNN embedding system. Through experiments on speakers in the wild (SITW) corpus, we show that the inclusion of total variability layer in a DNN embedding system provides around 20% relative improvement in equal error rate performance.

Noise Robust Speaker Recognition Based on Adaptive Frame Weighting in GMM for i-Vector Extraction

Even though speaker recognition has gained significant progress in recent years, its performance is known to be deteriorated severely with the existence of strong background noises. Inspired by a recently proposed clean-frame selection approach, this work investigates a relatively elegant weighting method when computing the Baum-Welch statistics of Gaussian mixture models (GMMs) in i-vector extraction. By introducing weighting parameters to the frames of enrollment/testing utterances, the optimization problem is redefined and solved. New updating rules are derived by incorporating weights to the computation of posterior probabilities, mean vectors, and covariance matrices of the GMM. The experiments conducted on the Speakers in the Wild (SITW) database show that the proposed algorithm has significantly improved the performance of i-vector-based speaker recognition systems in noisy environments. Compared with the GMM i-vector baseline, the equal error rate is reduced from 5.75 to 4.72 and the minimum value of cost function (C det min ) is reduced from 0.4825 to 0.4505. Slight but significant superiority is also observed over the method with an additional feature enhancement frontend by using deep neural networks.