Deep feature for text-dependent speaker verification

Liu, Yuan; Ye, Qian; Chen, Nanxin; Fu, Tianfan; Zhang, Ya; Yu, Kai

doi:10.1016/j.specom.2015.07.003

Cited by 162 publications

(132 citation statements)

References 35 publications

Supporting

Mentioning

131

Contrasting

Unclassified

Order By: Relevance

“…In [8], experimental results show that adopting multi-task learning enhances d-vector based speaker verification systems. Therefore, it can be said that multi-task learning provides an opening that allows humans to give additional information to neural networks when the additional information is considered helpful.…”

Section: Multi-task Learning In D-vector Based Speaker Verification (mentioning

confidence: 99%

“…This paper expands on the above systems and proposes a d-vector based speaker verification system in which raw waveform signals are used as input to the speaker identifier DNN. The concept of multi-task learning is applied throughout this paper following study results that indicate that multi-task learning enables speaker identifier DNN to extract more general and robust speaker identity representation [8] [9].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

D-vector based speaker verification system using Raw Waveform CNN

Jung¹,

Heo²,

Yang³

et al. 2018

Proceedings of the 2017 International Seminar on Artificial Intelligence, Networking and Information Technology (ANIT 2017)

View full text Add to dashboard Cite

Section: Multi-task Learning In D-vector Based Speaker Verification (mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

D-vector based speaker verification system using Raw Waveform CNN

Jung¹,

Heo²,

Yang³

et al. 2018

Proceedings of the 2017 International Seminar on Artificial Intelligence, Networking and Information Technology (ANIT 2017)

View full text Add to dashboard Cite

“…Many experts have done a lot in the corresponding studies [1][2][3] of speaker recognition. At present, the most popular features of speaker recognition are MFCC and LPCC [4][5][6]. In terms of recognition methods, vector quantization [7] , Gaussian Mixture Model(GMM) [7][8] and Hidden Markov Models(HMM) had gradually been applied in the field of speaker recognition.…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, support vector machine (SVM) [8] has been widely used due to the use of superscalar features. What's more, the neural network algorithm especially deep learning network algorithm [6] has now become the new direction of speaker recognition algorithm.…”

Section: Introductionmentioning

confidence: 99%

Robust Speaker Recognition Algorithm

Hao¹,

Chen²,

Wang³

et al. 2017

Proceedings of the International Conference on Communication and Electronic Information Engineering (CEIE 2016)

View full text Add to dashboard Cite

The accuracy of speaker recognition algorithm would be decreased greatly due to the noise issues. According to noisy environment, a new robust speaker recognition algorithm is proposed in this paper. After Mel-frequency Cepstral Coefficient (MFCC) feature extraction, the features are calibrated with half rised-sine function. Then the features are processed by feature normalization, feature folding and feature mapping. A method of combining BP neural network(NN) with Gaussian mixture model(GMM) is proposed to improve the recognition accuracy and robustness of the model. The neural network works in the probability space of GMM gathers the interactive information between different speakers. The experiment result proves that the proposed algorithm shows more accuracy and robustness.

show abstract

“…It is also worth noting that recently different methods have been proposed to incorporate Deep Neural Networks (DNNs) into the context of speaker recognition. In [18,19], DNNs were used to extract an enriched vector representation of speakers for textdependent speaker verification. In [20], DNNs were employed to extract a more discriminative vector from i-vector.…”

Section: Introductionmentioning

confidence: 99%

From Features to Speaker Vectors by means of Restricted Boltzmann Machine Adaptation

Safari¹,

Ghahabi²,

Hernando³

2016

The Speaker and Language Recognition Workshop (Odyssey 2016)

View full text Add to dashboard Cite

Restricted Boltzmann Machines (RBMs) have shown success in different stages of speaker recognition systems. In this paper, we propose a novel framework to produce a vector-based representation for each speaker, which will be referred to as RBMvector. This new approach maps the speaker spectral features to a single fixed-dimensional vector carrying speaker-specific information. In this work, a global model, referred to as Universal RBM (URBM), is trained taking advantage of RBM unsupervised learning capabilities. Then, this URBM is adapted to the data of each speaker in the development, enrolment and evaluation datasets. The network connection weights of the adapted RBMs are further concatenated and subject to a whitening with dimension reduction stage to build the speaker vectors. The evaluation is performed on the core test condition of the NIST SRE 2006 database, and it is shown that RBM-vectors achieve 15% relative improvement in terms of EER compared to i-vectors using cosine scoring. The score fusion with i-vector attains more than 24% relative improvement. The interest of this result for score fusion yields on the fact that both vectors are produced in an unsupervised fashion and can be used instead of i-vector/PLDA approach, when no data label is available. Results obtained for RBM-vector/PLDA framework is comparable with the ones from i-vector/PLDA. Their score fusion achieves 14% relative improvement compared to i-vector/PLDA.

show abstract

Deep feature for text-dependent speaker verification

Cited by 162 publications

References 35 publications

D-vector based speaker verification system using Raw Waveform CNN

D-vector based speaker verification system using Raw Waveform CNN

Robust Speaker Recognition Algorithm

From Features to Speaker Vectors by means of Restricted Boltzmann Machine Adaptation

Contact Info

Product

Resources

About