Jenthe Thienpondt

Jenthe Thienpondt, Kris Demuynck

Automatic Speech Recognition (ASR) systems are known to exhibit difficulties when transcribing children's speech. This can mainly be attributed to the absence of large children's speech corpora to train robust ASR models and the resulting domain mismatch when decoding children's speech with systems trained on adult data. In this paper, we propose multiple enhancements to alleviate these issues. First, we propose a data augmentation technique based on the source-filter model of speech to close the domain gap between adult and children's speech. This enables us to leverage the data availability of adult speech corpora by making these samples perceptually similar to children's speech. Second, using this augmentation strategy, we apply transfer learning on a Transformer model pre-trained on adult data. This model follows the recently introduced XLS-R architecture, a wav2vec 2.0 model pre-trained on several cross-lingual adult speech corpora to learn general and robust acoustic frame-level representations. Adopting this model for the ASR task using adult data augmented with the proposed source-filter warping strategy and a limited amount of in-domain children's speech significantly outperforms previous state-of-the-art results on the PF-STAR British English Children's Speech corpus with a 4.86% WER on the official test set.

Jenthe Thienpondt, Brecht Desplanques, Kris Demuynck

This paper contains a post-challenge performance analysis on cross-lingual speaker verification of the IDLab submission to the VoxCeleb Speaker Recognition Challenge 2021 (VoxSRC-21). We show that current speaker embedding extractors consistently underestimate speaker similarity in within-speaker cross-lingual trials. Consequently, the typical training and scoring protocols do not put enough emphasis on the compensation of intra-speaker language variability. We propose two techniques to increase cross-lingual speaker verification robustness. First, we enhance our previously proposed Large-Margin Fine-Tuning (LM-FT) training stage with a mini-batch sampling strategy which increases the amount of intra-speaker cross-lingual samples within the mini-batch. Second, we incorporate language information in the logistic regression calibration stage. We integrate quality metrics based on soft and hard decisions of a VoxLingua107 language identification model. The proposed techniques result in a 11.7% relative improvement over the baseline model on the VoxSRC-21 test set and contributed to our third place finish in the corresponding challenge.

Jenthe Thienpondt, Brecht Desplanques, Kris Demuynck

This technical report describes the IDLab submission for track 1 and 2 of the VoxCeleb Speaker Recognition Challenge 2021 (VoxSRC-21). This speaker verification competition focuses on short duration test recordings and cross-lingual trials. Currently, both Time Delay Neural Networks (TDNNs) and ResNets achieve state-of-the-art results in speaker verification. We opt to use a system fusion of hybrid architectures in our final submission. An ECAPA-TDNN baseline is enhanced with a 2D convolutional stem to transfer some of the strong characteristics of a ResNet based model to this hybrid CNN-TDNN architecture. Similarly, we incorporate absolute frequency positional information in the SE-ResNet architectures. All models are trained with a special mini-batch data sampling technique which constructs mini-batches with data that is the most challenging for the system on the level of intra-speaker variability. This intra-speaker variability is mainly caused by differences in language and background conditions between the speaker's utterances. The cross-lingual effects on the speaker verification scores are further compensated by introducing a binary cross-linguality trial feature in the logistic regression based system calibration. The final system fusion with two ECAPA CNN-TDNNs and three SE-ResNets enhanced with frequency positional information achieved a third place on the VoxSRC-21 leaderboard for both track 1 and 2 with a minDCF of 0.1291 and 0.1313 respectively.

Jenthe Thienpondt, Brecht Desplanques, Kris Demuynck

In this technical report we describe the IDLAB top-scoring submissions for the VoxCeleb Speaker Recognition Challenge 2020 (VoxSRC-20) in the supervised and unsupervised speaker verification tracks. For the supervised verification tracks we trained 6 state-of-the-art ECAPA-TDNN systems and 4 Resnet34 based systems with architectural variations. On all models we apply a large margin fine-tuning strategy, which enables the training procedure to use higher margin penalties by using longer training utterances. In addition, we use quality-aware score calibration which introduces quality metrics in the calibration system to generate more consistent scores across varying levels of utterance conditions. A fusion of all systems with both enhancements applied led to the first place on the open and closed supervised verification tracks. The unsupervised system is trained through contrastive learning. Subsequent pseudo-label generation by iterative clustering of the training embeddings allows the use of supervised techniques. This procedure led to the winning submission on the unsupervised track, and its performance is closing in on supervised training.

Jenthe Thienpondt, Brecht Desplanques, Kris Demuynck

In this paper we propose and analyse a large margin fine-tuning strategy and a quality-aware score calibration in text-independent speaker verification. Large margin fine-tuning is a secondary training stage for DNN based speaker verification systems trained with margin-based loss functions. It enables the network to create more robust speaker embeddings by enabling the use of longer training utterances in combination with a more aggressive margin penalty. Score calibration is a common practice in speaker verification systems to map output scores to well-calibrated log-likelihood-ratios, which can be converted to interpretable probabilities. By including quality features in the calibration system, the decision thresholds of the evaluation metrics become quality-dependent and more consistent across varying trial conditions. Applying both enhancements on the ECAPA-TDNN architecture leads to state-of-the-art results on all publicly available VoxCeleb1 test sets and contributed to our winning submissions in the supervised verification tracks of the VoxCeleb Speaker Recognition Challenge 2020.

Jenthe Thienpondt, Brecht Desplanques, Kris Demuynck

In this paper we describe the top-scoring IDLab submission for the text-independent task of the Short-duration Speaker Verification (SdSV) Challenge 2020. The main difficulty of the challenge exists in the large degree of varying phonetic overlap between the potentially cross-lingual trials, along with the limited availability of in-domain DeepMine Farsi training data. We introduce domain-balanced hard prototype mining to fine-tune the state-of-the-art ECAPA-TDNN x-vector based speaker embedding extractor. The sample mining technique efficiently exploits speaker distances between the speaker prototypes of the popular AAM-softmax loss function to construct challenging training batches that are balanced on the domain-level. To enhance the scoring of cross-lingual trials, we propose a language-dependent s-norm score normalization. The imposter cohort only contains data from the Farsi target-domain which simulates the enrollment data always being Farsi. In case a Gaussian-Backend language model detects the test speaker embedding to contain English, a cross-language compensation offset determined on the AAM-softmax speaker prototypes is subtracted from the maximum expected imposter mean score. A fusion of five systems with minor topological tweaks resulted in a final MinDCF and EER of 0.065 and 1.45% respectively on the SdSVC evaluation set.

Brecht Desplanques, Jenthe Thienpondt, Kris Demuynck

Current speaker verification techniques rely on a neural network to extract speaker representations. The successful x-vector architecture is a Time Delay Neural Network (TDNN) that applies statistics pooling to project variable-length utterances into fixed-length speaker characterizing embeddings. In this paper, we propose multiple enhancements to this architecture based on recent trends in the related fields of face verification and computer vision. Firstly, the initial frame layers can be restructured into 1-dimensional Res2Net modules with impactful skip connections. Similarly to SE-ResNet, we introduce Squeeze-and-Excitation blocks in these modules to explicitly model channel interdependencies. The SE block expands the temporal context of the frame layer by rescaling the channels according to global properties of the recording. Secondly, neural networks are known to learn hierarchical features, with each layer operating on a different level of complexity. To leverage this complementary information, we aggregate and propagate features of different hierarchical levels. Finally, we improve the statistics pooling module with channel-dependent frame attention. This enables the network to focus on different subsets of frames during each of the channel's statistics estimation. The proposed ECAPA-TDNN architecture significantly outperforms state-of-the-art TDNN based systems on the VoxCeleb test sets and the 2019 VoxCeleb Speaker Recognition Challenge.