A Novel Scheme for Speaker Recognition Using a Phonetically-Aware Deep Neural Network

Lei, Yun; Scheffer, Nicolas; Ferrer, Luciana; McLaren, Mitchell

doi:10.21236/ada613971

Cited by 289 publications

(300 citation statements)

References 11 publications

Supporting

Mentioning

296

Contrasting

Order By: Relevance

“…The focus of the speaker recognition community is mainly concerned with the monaural or single-channel case as speaker recognition has traditionally been applied to telephone speech. Major advances on increasing the robustness of speaker recognition include the introduction of the i-vector framework based on Gaussian mixture models (GMM/i-vector) [1], the Probabilistic Linear Discriminant Analysis (PLDA) back-end [2], Deep Neural Networks (DNNs) replacing the GMM component [3], and recently introduced x-vectors [4] for speaker embedding. X-vectors are of particular interest due to their use of inexpensive data augmentation for increasing robustness and utilization of the algorithms associated with i-vectors, e. g. PLDA for scoring.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Based Multi-Channel Speaker Recognition in Noisy and Reverberant Environments

Taherian

Wang

2019

Interspeech 2019

View full text Add to dashboard Cite

Despite successful applications of multi-channel signal processing in robust automatic speech recognition (ASR), relatively little research has been conducted on the effectiveness of such techniques in the robust speaker recognition domain. This paper introduces time-frequency (T-F) maskingbased beamforming to address text-independent speaker recognition in conditions where strong diffuse noise and reverberation are both present. We examine various masking-based beamformers, such as parameterized multi-channel Wiener filter, generalized eigenvalue (GEV) beamformer and minimum variance distortion-less response (MVDR) beamformer, and evaluate their performance in terms of speaker recognition accuracy for i-vector and x-vector based systems. In addition, we present a different formulation for estimating steering vectors from speech covariance matrices. We show that rank-1 approximation of a speech covariance matrix based on generalized eigenvalue decomposition leads to the best results for the masking-based MVDR beamformer. Experiments on the recently introduced NIST SRE 2010 retransmitted corpus show that the MVDR beamformer with rank-1 approximation provides an absolute reduction of 5.55% in equal error rate compared to a standard masking-based MVDR beamformer.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep Learning Based Multi-Channel Speaker Recognition in Noisy and Reverberant Environments

Taherian

Wang

2019

Interspeech 2019

View full text Add to dashboard Cite

show abstract

“…This method is for classification only. The deep neural network is a standard feed‐forward neural network that is both much larger and much deeper than traditional neural networks (Lei, Scheffer, Ferrer, & Mclaren, ) and it has recently led to significant improvement in countless areas of machine learning (Gu & Rigazio, ).…”

Section: Methodsmentioning

confidence: 99%

In silico prediction of drug‐induced rhabdomyolysis with machine‐learning models and structural alerts

Cui

Liu

Zhang

et al. 2019

J of Applied Toxicology

View full text Add to dashboard Cite

Drug-induced rhabdomyolysis (DIR) is a serious adverse reaction and can be fatal. In the present study, we focused on the modeling and understanding of the molecular basis of DIR of small molecule drugs. A series of machine-learning models were developed using an Online Chemical Modeling Environment platform with a diverse dataset. A total of 80 machine-learning models were generated. Based on the topperforming individual models, a consensus model was also developed. The consensus model was available at https://ochem.eu/model/32214665, and the individual models can be accessed with the corresponding model IDs on the website. Furthermore, we also analyzed the difference of distributions of eight key physicochemical properties between rhabdomyolysis-inducing drugs and non-rhabdomyolysisinducing drugs. Finally, structural alerts responsible for DIR were identified from fragments of the Klekota-Roth fingerprints.

show abstract

“…It is written as = + where μ is mean super-vector, is low rank total variability matrix and is a low rank vector referred to as i-vector. PLDA model is then employed to generate verification scores by comparing i-vectors from different utterances [7]. Any model that can provide posteriors of K-classes per frame other than GMM can be used for i-vector calculation.…”

Section: A Gmm-ubm and I-vector Based Systemmentioning

confidence: 99%

Speaker Verification Using Deep Neural Networks: A Review

Irum¹,

Ahmad²

2019

IJMLC

View full text Add to dashboard Cite

Speaker verification involves examining the speech signal to authenticate the claim of a speaker as true or false. Deep neural networks are one of the successful implementation of complex non-linear models to learn unique and invariant features of data. They have been employed in speech recognition tasks and have shown their potential to be used for speaker recognition also. In this study, we investigate and review Deep Neural Network (DNN) techniques used in speaker verification systems. DNN are used from extracting features to complete end-to-end system for speaker verification. They are generally used to extract speaker-specific representations, for which the network is trained using speaker data in training phase. Speaker representation depends on the type of the model, the representation level, and the model training loss. Usually deep learning is crux of attention in computer vision community for various tasks and we believe that a comprehensive review of current state-of-the-art in deep learning for speaker verification summarize the utilization of these approaches for readers in speech processing community. Index Terms-Feature extraction, bottleneck features, deep features, end-to-end systems.

show abstract

A Novel Scheme for Speaker Recognition Using a Phonetically-Aware Deep Neural Network

Cited by 289 publications

References 11 publications

Deep Learning Based Multi-Channel Speaker Recognition in Noisy and Reverberant Environments

Deep Learning Based Multi-Channel Speaker Recognition in Noisy and Reverberant Environments

In silico prediction of drug‐induced rhabdomyolysis with machine‐learning models and structural alerts

Speaker Verification Using Deep Neural Networks: A Review

Contact Info

Product

Resources

About