A Light Convolutional GRU-RNN Deep Feature Extractor for ASV Spoofing Detection

Gomez-Alanis, Alejandro; Peinado, Antonio M.; González, José A.; Gómez, Ángel M.

doi:10.21437/interspeech.2019-2212

Cited by 62 publications

(37 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also used the Light Convolutional Gated Recurrent Neural Network (LC-GRNN) that we proposed in our previous works [9], [19]. It was one of the ten top performing single systems of the ASVspoof 2019 challenge [8].…”

Section: ) Light Convolutional Gated Recurrent Neural Networkmentioning

confidence: 99%

“…The embeddings extracted from the utterances were finally processed by a classifier, which produces a score per utterance, indicating whether the utterance is genuine or spoofed. Based on the results from our previous works [9], [19], we used a probabilistic linear discriminant analysis (PLDA). We also applied a posterior normalization of the scores.…”

Section: ) Final Classifiermentioning

confidence: 99%

“…Spoofing detection or presentation attack detection (PAD in ISO/IEC 30107 nomenclature [5]) for ASV has become a hot research topic in recent years as evidenced by the organization of several evaluation campaigns (challenges) in this specific topic: (i) ASVspoof 2015 [6], which focused on logical access (LA) attacks (TTS and VC); (ii) ASVspoof 2017 [7], which focused on physical access (PA) attacks (replay attacks) under noisy environments; and (iii) ASVspoof 2019 [8], which addressed both the detection of LA attacks generated with the latest TTS and VC technologies, and simulated replay attacks under different reverberant acoustic VOLUME X, 2020 conditions. One of the main conclusions withdrawn from these challenges is that the use of deep neural networks (DNNs) for the extraction of spoofing-aware embedding vectors outperforms other conventional approaches for ASV anti-spoofing [9]- [12].…”

Section: Introductionmentioning

confidence: 99%

“…Three key points are important for building a DNN-based antispoofing system with generalization capabilities: (i) architecture, (ii) input features, and (iii) loss function. Multiple types of DNN architectures have been explored, such as feedforward DNN [16], convolutional neural network (CNN) [16], [17], recurrent neural network (RNN) [17], [18], gated recurrent neural network (GRCNN) [19], light convolutional gated recurrent neural network (LC-GRNN) [9], light convolutional neural network (LCNN) [20], central difference convolutional network (CDCN) [14], etc. Also, a wide range of features have been proposed to train these models, such as spectrogram [12], linear frequency cepstral coefficients (LFCC) [21], constant Q cepstral coefficients (CQCC) [22], raw speech samples [23], local similar pattern (LSP) features [15], signal-to-noise mask (SNM) [19] features, etc.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Kernel Density Estimation Based Loss Function and its Application to ASV-Spoofing Detection

2020

Self Cite

View full text Add to dashboard Cite

Biometric systems are exposed to spoofing attacks which may compromise their security, and voice biometrics, also known as automatic speaker verification (ASV), is no exception. Replay, synthesis and voice conversion attacks cause false acceptances that can be detected by anti-spoofing systems. Recently, deep neural networks (DNNs) which extract embedding vectors have shown superior performance than conventional systems in both ASV and anti-spoofing tasks. In this work, we develop a new concept of loss function for training DNNs which is based on kernel density estimation (KDE) techniques. The proposed loss functions estimate the probability density function (pdf) of every training class in each mini-batch, and compute a log likelihood matrix between the embedding vectors and pdfs of all training classes within the mini-batch in order to obtain the KDE-based loss. To evaluate our proposal for spoofing detection, experiments were carried out on the recent ASVspoof 2019 corpus, including both logical and physical access scenarios. The experimental results show that training a DNN based anti-spoofing system with our proposed loss functions clearly outperforms the performance of the same system being trained with other well-known loss functions. Moreover, the results also show that the proposed loss functions are effective for different types of neural network architectures.

show abstract

Section: ) Light Convolutional Gated Recurrent Neural Networkmentioning

confidence: 99%

Section: ) Final Classifiermentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Kernel Density Estimation Based Loss Function and its Application to ASV-Spoofing Detection

2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…In which, the corpus has three subset, train, development, and evaluation set. According to ASVspoof 2019 challenge rule, tandem detection cost function (t-DCF) [56] and EER are used as the primary and secondary metric, respectively, which is the same as the previous works [57][58][59][60][61][62][63][64]. Table 10 gives the experimental results on ASVspoof 2019 physical access development set using dynamic features of CMOC and CVOC.…”

Section: Database Introduction and Evaluation Metricmentioning

confidence: 99%

Discriminative features based on modified log magnitude spectrum for playback speech detection

Yang

Ren³

et al. 2020

J AUDIO SPEECH MUSIC PROC.

View full text Add to dashboard Cite

In order to improve the performance of hand-crafted features to detect playback speech, two discriminative features, constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients, are proposed for playback speech detection in this work. They rely on our findings that variance-based modified log magnitude spectrum and mean-based modified log magnitude spectrum can enhance the discriminative power between genuine speech and playback speech. Then constant-Q variance-based octave coefficients (constant-Q mean-based octave coefficients) can be obtained by combining variance-based modified log magnitude spectrum (mean-based modified log magnitude spectrum), octave segmentation, and discrete cosine transform. Finally, constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients are evaluated on ASVspoof 2017 corpus version 2.0 and ASVspoof 2019 physical access, respectively. Experimental results show that variance-based modified log magnitude spectrum and mean-based modified log magnitude spectrum can produce discriminative features toward playback speech. Further results on the two databases show that constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients can perform better than some common features, such as mel frequency cepstral coefficients and constant-Q cepstral coefficients.

show abstract

Attention Network with GMM Based Feature for ASV Spoofing Detection

Lei

Yang

et al. 2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

A Light Convolutional GRU-RNN Deep Feature Extractor for ASV Spoofing Detection

Cited by 62 publications

References 0 publications

A Kernel Density Estimation Based Loss Function and its Application to ASV-Spoofing Detection

A Kernel Density Estimation Based Loss Function and its Application to ASV-Spoofing Detection

Discriminative features based on modified log magnitude spectrum for playback speech detection

Attention Network with GMM Based Feature for ASV Spoofing Detection

Contact Info

Product

Resources

About