CN-CELEB: a challenging Chinese speaker recognition dataset

Fan, Yue; Kang, Jiawen; Li, Lantian; Li, Kaicheng; Chen, Haolin; Cao, Shuai; Zhang, Pengyuan; Zhou, Ziya; Cai, Yunqi; Wang, Dong

doi:10.48550/arxiv.1911.01799

Cited by 3 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Three datasets were used in our experiments: VoxCeleb [20,21], SITW [22] and CNCeleb [23]. VoxCeleb was used to train models (x-vector, PLDA, LDA, subspace DNF), while the other two were used for performance evaluation.…”

Section: Data and Settingmentioning

confidence: 99%

Deep generative LDA

Cai,

Wang

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Linear discriminant analysis (LDA) is a popular tool for classification and dimension reduction. Limited by its linear form and the underlying Gaussian assumption, however, LDA is not applicable in situations where the data distribution is complex. Recently, we proposed a discriminative normalization flow (DNF) model. In this study, we reinterpret DNF as a deep generative LDA model, and study its properties in representing complex data. We conducted a simulation experiment and a speaker recognition experiment. The results show that DNF and its subspace version are much more powerful than the conventional LDA in modeling complex data and retrieving low-dimensional representations.

show abstract

Section: Data and Settingmentioning

confidence: 99%

Deep generative LDA

Cai,

Wang

2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Notable works include masking [1] and mapping [2] based approach, Speech Enhancement Generative Adversarial Network (SEGAN) [3], Deep Feature Loss (DFL) [4], end-to-end metric optimization [5], and Transformer based approach [6,7]. Meanwhile, an active research exists in the robustness of Speaker Verification (SV) systems [8,9,10,11]. Another reason for interest in speech enhancement arises from the notion that it is considered as a modern solution to improve noise robustness in SV systems [10,12,13].…”

Section: Introductionmentioning

confidence: 99%

“…one trained with (noisy) data augmentations is superior. Training data choice is important to us because we focus on BabyTrain and large "in the wild" public data releases such as SITW [19], VoxCeleb [23], and CN-Celeb [9] do not explicitly account for children speech.…”

Section: Introductionmentioning

confidence: 99%

Analysis of Deep Feature Loss based Enhancement for Speaker Verification

Kataria¹,

Nidadavolu²,

Villalba³

et al. 2020

Preprint

View full text Add to dashboard Cite

Data augmentation is conventionally used to inject robustness in Speaker Verification systems. Several recently organized challenges focus on handling novel acoustic environments. Deep learning based speech enhancement is a modern solution for this. Recently, a study proposed to optimize the enhancement network in the activation space of a pre-trained auxiliary network. This methodology, called deep feature loss, greatly improved over the state-of-the-art conventional x-vector based system on a children speech dataset called BabyTrain. This work analyzes various facets of that approach and asks few novel questions in that context. We first search for optimal number of auxiliary network activations, training data, and enhancement feature dimension. Experiments reveal the importance of Signal-to-Noise Ratio filtering that we employ to create a large, clean, and naturalistic corpus for enhancement network training. To counter the "mismatch" problem in enhancement, we find enhancing front-end (x-vector network) data helpful while harmful for the back-end (Probabilistic Linear Discriminant Analysis (PLDA)). Importantly, we find enhanced signals contain complementary information to original. Established by combining them in front-end, this gives ~40% relative improvement over the baseline. We also do an ablation study to remove a noise class from x-vector data augmentation and, for such systems, we establish the utility of enhancement regardless of whether it has seen that noise class itself during training. Finally, we design several dereverberation schemes to conclude ineffectiveness of deep feature loss enhancement scheme for this task.

show abstract

Deep Normalization for Speaker Vectors

Cai¹,

Wang³

et al. 2020

Preprint

View full text Add to dashboard Cite

CN-CELEB: a challenging Chinese speaker recognition dataset

Cited by 3 publications

References 17 publications

Deep generative LDA

Deep generative LDA

Analysis of Deep Feature Loss based Enhancement for Speaker Verification

Deep Normalization for Speaker Vectors

Contact Info

Product

Resources

About