The DKU System for the Speaker Recognition Task of the 2019 VOiCES from a Distance Challenge

Abstract

In this paper, we present the DKU system for the speaker recognition task of the VOiCES from a distance challenge 2019. We investigate the whole system pipeline for the far-field speaker verification, including data pre-processing, short-term spectral feature representation, utterance-level speaker modeling, backend scoring, and score normalization. Our best single system employs a residual neural network trained with angular softmax loss. Also, the weighted prediction error algorithms can further improve performance. It achieves 0.3668 minDCF and 5.58% EER on the evaluation set by using a simple cosine similarity scoring. Finally, the submitted primary system obtains 0.3532 minDCF and 4.96% EER on the evaluation set.

Multi-Channel Training for End-to-End Speaker Recognition Under Reverberant and Noisy Environment

“…After training, the speaker embedding adopts cosine similarity for scoring. In the deep speaker embedding system with ResNet + GAP setting, a cosine similarity backend is sufficient to achieve good performance [4,22]. For training data, the original clean speech is used to train the deep speaker embedding system.…”

“…DNN based denoising methods for single-channel speech enhancement [10,11,12,13] and beamforming for multi-channel speech enhancement [9,14,15] have also been investigated for ASV under complex environment. At feature level, sub-band Hilbert envelopes based features [16,17,18], warped minimum variance distortionless response (MVDR) cepstral coefficients [19], blind spectral weighting (BSW) based features [20], power-normalized cepstral coefficients (PNCC) [21,22] and DNN bottleneck features [23] have been applied to ASV system to suppress the adverse impacts of reverberation and noise. At the model level, reverberation matching with multi-condition training models have been successfully employed within the universal background model (UBM) or i-vector based front-end systems [24,25].…”

“…In back-end modeling, multicondition training of probabilistic linear discriminant analysis (PLDA) models was employed in i-vector system [27]. The robustness of deep speaker embeddings for far-field speech has also been investigated in [22,28]. Finally, at the score level, score normalization [24] and multi-channel score fusion [29,30] have been applied in far-field ASV system to improve the robustness.…”

HI-MIA: A Far-Field Text-Dependent Speaker Verification Database and the Baselines

“…The superiority of deep speaker embedding systems has been shown in text-independent speaker recognition for closed talking [21,22] and far-field scenarios [24,25]. In this paper, we Figure 2: Gender and age distribution adopt the deep speaker embedding system, which is initially designed for the text-independent speaker verification, as baseline.…”

Far-Field End-to-End Text-Dependent Speaker Verification Based on Mixed Training Data with Transfer Learning and Enrollment Data Augmentation

“…In backend modeling, multi-condition training of probabilistic linear discriminant analysis (PLDA) models was employed in i-vector system [25]. The robustness of deep speaker embeddings for far-field text-independent speech has also been investigated in [26,27]. Finally, at the score level, score normalization [22] and multi-channel score fusion [28] have been applied in farfield ASV system to improve the robustness.…”