Replay attack detection with auditory filter-based relative phase features

Abstract

There are many studies on detecting human speech from artificially generated speech and automatic speaker verification (ASV) that aim to detect and identify whether the given speech belongs to a given speaker. Recent studies demonstrate the success of the relative phase (RP) feature in speaker recognition/verification and the detection of synthesized speech and converted speech. However, there are few studies that focus on the RP feature for replay attack detection. In this paper, we improve the discriminating ability of the RP feature by proposing two new auditory filter-based RP features for replay attack detection. The key idea is to integrate the advantage of RP-based features in signal representation with the advantage of two auditory filter-based RP features. For the first proposed feature, we apply a Mel-filter bank to convert the signal representation of conventional RP information from a linear scale to a Mel scale, where the modified representation is called the Mel-scale RP feature. For the other proposed feature, a gammatone filter bank is applied to scale the RP information, where the scaled RP feature is called the gammatone-scale RP feature. These two proposed phase-based features are implemented to achieve better performance than a conventional RP feature because of the scale resolution and. In addition to the use of individual Mel/gammatone-scale RP features, a combination of the scores of these proposed RP features and a standard magnitude-based feature, that is, the constant Q transform cepstral coefficient (CQCC), is also applied to further improve the reliable detection decision. The effectiveness of the proposed Mel-scale RP feature, gammatone-scale RP feature, and their combination are evaluated using the ASVspoof 2017 dataset. On the evaluation dataset, our proposed methods demonstrate significant improvement over the existing feature and baseline CQCC feature. The combination of the CQCC and gammatone-scale RP provides the best performance compared with an individual baseline feature and other combination methods.

Replay Attack Detection Using Linear Prediction Analysis-Based Relative Phase Features

“…By comparing the proposed LPR-RP/LPAES-RP with the LPR-MFCC, we noticed that the phase-based feature provided encouraging performance for replay attack detection. Likewise, the LFMGDCC in [41] and DCT-linear-RPS in [41], IFCC [33], mel-RP in [36], and gammatone-RP in [39] were confirmed to be efficient phase features under unseen evaluation. However, the LFMGDCC feature may lose some representation in the vocal source information of the given speech due to the enhancements of the envelope of the short-time speech spectrum, and the phase shift variation in the DCT-Linear-RPS is not normalized by cutting positions.…”

“…Although the RP has been successfully implemented for spoofing attack detection, its discriminating ability for replay attack detection can be further improved using a frequency resolution. In our previous work [36]- [39], mel/inverted mel-scale filterbank, linear-scale filterbank, attention-based adaptive filterbank, and gammatone-scale filterbank were applied to convert the RP information from the original linear scale to new scales (such as mel-scale, inverted mel-scale, and gammatone-scale), where the modified RP information is called mel-scale RP [37], inverted mel-scale RP (IMel-RP) [36], linear-scale RP (linear-RP) [36], adaptive scale RP (ARP) [38], gammatone-scale RP (gamatone-RP) [39], respectively. The results demonstrated that the mel-RP, ARP, and gammatone-RP outperformed the original RP feature due to the frequency resolution of the filterbank.…”

“…The results demonstrated that the mel-RP, ARP, and gammatone-RP outperformed the original RP feature due to the frequency resolution of the filterbank. Moreover, additional improvement could be obtained by combining the mel-RP/ARP/gammatone-RP with MFCC/IMFCC/CQCC at score level, especially in the combination of the gammatone-RP and CQCC yielding the best result compared with the related auditory filterbank-based RP features [36]- [39]. However, the modified phase information using the filterbank may lose some information for score combination with magnitude-based features.…”

EarArray: Defending against DolphinAttack via Acoustic Attenuation

“…In the antispoofing competition ASV2017 [31], Witkowski, et al [32] pointed out that replay attacks can be detected by analyzing the high-frequency band of the replayed recordings. Zeyan et al [33] improved the discriminating ability of the relative phase (RP) features by proposing two new auditory filterbased RP features for replay attack detection. To detect the remote attaker, Lee et al [12] proposed a sonar-based liveness detection system.…”

A multi-branch ResNet with discriminative features for detection of replay speech signals

“…To utilize the phase information, Tom et al [18] used the group delay function (GD) in replay detection. Oo et al [19] introduced the relative phase (RP) feature and further extended it in the Mel-scale (Mel-RP) and the gammatone-scale (Gamma-RP). Phapatanaburi et al [20] proposed to extract RP based on the linear prediction analysis (LPA), which extracted RP on the residual signal of LPA.…”