Decision-level Feature Switching as a Paradigm for Replay Attack Detection

IEICE Trans. Inf. & Syst.

Ren

2020

Under the framework of traditional power spectrum based feature extraction, in order to extract more discriminative information for playback attack detection, this paper proposes a feature by making use of deep neural network to describe the nonlinear relationship between power spectrum and discriminative information. Namely, constant-Q deep coefficients (CQDC). It relies on constant-Q transform, deep neural network and discrete cosine transform. In which, constant-Q transform is used to convert signal from the time domain into the frequency domain because it is a long-term transform that can provide more frequency detail, deep neural network is used to extract more discriminative information to discriminate playback speech from genuine speech and discrete cosine transform is used to decorrelate among the feature dimensions. ASVspoof 2017 corpus version 2.0 is used to evaluate the performance of CQDC. The experimental results show that CQDC outperforms the existing power spectrum obtained from constant-Q transform based features, and equal error can reduce from 19.18% to 51.56%. In addition, we found that discriminative information of CQDC hides in all frequency bins, which is different from commonly used features.

Section: Comparison With Some Known Systemsmentioning

confidence: 78%

Section: Comparison With Different Dimensionsmentioning

confidence: 99%

Constant-Q Deep Coefficients for Playback Attack Detection

IEICE Trans. Inf. & Syst.

Ren

2020

“…• Discrete Fourier transform (DFT) based features: which include Mel frequency cepstral coefficients (MFCC) [4,13,36], mel filterbank slope [10], linear filterbak slope [10], and Q-log domain DFT-based mean normalized log spectral [42]. • Variable length energy separation algorithm (VESA)-based features: which include instantaneous frequency cosine coefficients based on VESA [6] and instantaneous amplitude cosine coefficients based on VESA [43].…”

Section: Related Workmentioning

confidence: 99%

“…Since the ASVspoof 2017 challenge [1,2], more and more researchers begin to focus on playback speech detection [3][4][5][6][7][8][9][10]. Similar to many speech signal processing systems, most of all playback speech detection systems usually consist of front-end feature and back-end classifier [11][12][13][14][15][16][17][18].…”

Section: Introductionmentioning

confidence: 99%

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

J AUDIO SPEECH MUSIC PROC.

Ren³

et al. 2020

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.

“…On the other hand, the need to select informative feature representations arises when building a system with a growing number of features. As a consequence, choosing fewer but sufficient, complementary features is of supreme importance [10]. In this paper, we focus on resolving the two challenges mentioned above, especially the first one.…”

Section: Introductionmentioning

confidence: 99%

Dynamically Mitigating Data Discrepancy with Balanced Focal Loss for Replay Attack Detection

Dou

et al. 2020

Preprint

It becomes urgent to design effective anti-spoofing algorithms for vulnerable automatic speaker verification systems due to the advancement of high-quality playback devices. Current studies mainly treat anti-spoofing as a binary classification problem between bonafide and spoofed utterances, while lack of indistinguishable samples makes it difficult to train a robust spoofing detector. In this paper, we argue that for antispoofing, it needs more attention for indistinguishable samples over easily-classified ones in the modeling process, to make correct discrimination a top priority. Therefore, to mitigate the data discrepancy between training and inference, we propose to leverage a balanced focal loss function as the training objective to dynamically scale the loss based on the traits of the sample itself. Besides, in the experiments, we select three kinds of features that contain both magnitude-based and phase-based information to form complementary and informative features. Experimental results on the ASVspoof2019 dataset demonstrate the superiority of the proposed methods by comparison between our systems and top-performing ones. Systems trained with the balanced focal loss perform significantly better than conventional cross-entropy loss. With complementary features, our fusion system with only three kinds of features outperforms other systems containing five or more complex single models by 22.5% for min-tDCF and 7% for EER, achieving a min-tDCF and an EER of 0.0124 and 0.55% respectively. Furthermore, we present and discuss the evaluation results on real replay data apart from the simulated ASVspoof2019 data, indicating that research for anti-spoofing still has a long way to go.