A comparison of features for synthetic speech detection
Open accessThe DOI matches its Crossref record; 36 of the 36 references it cites were found in a registry or an index.
Abstract
The performance of biometric systems based on automatic speaker recognition technology is severely degraded due to spoofing attacks with synthetic speech generated using different voice conversion (VC) and speech synthesis (SS) techniques. Various countermeasures are proposed to detect this type of at-tack, and in this context, choosing an appropriate feature extrac-tion technique for capturing relevant information from speech is an important issue. This paper presents a concise experi-mental review of different features for synthetic speech detec-tion task. A wide variety of features considered in this study include previously investigated features as well as some other potentially useful features for characterizing real and synthetic speech. The experiments are conducted on recently released ASVspoof 2015 corpus containing speech data from a large number of VC and SS technique. Comparative results using two different classifiers indicate that features representing spectral information in high-frequency region, dynamic information of speech, and detailed information related to subband characteris-tics are considerably more useful in detecting synthetic speech. Index Terms: anti-spoofing, ASVspoof 2015, feature extrac-tion, countermeasures
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“a significant body of research [64,96,97,98,99] has demonstrated that the performance of detection models largely depends on the choice of input features.”
From Audio Deepfake Detection: What Has Been Achieved and What Lies Ahead (2025) · cites this paper as [96]
In the citing paper
“other machine learning models, such as the Support Vector Machine and Logistic Regression, were widely studied in earlier research [96,134,135,136]. With the advancement of feature extraction techniques, these models continue to play a significant role in current studies [137,138,139].”
From Audio Deepfake Detection: What Has Been Achieved and What Lies Ahead (2025) · cites this paper as [96]
Citations
Cites
Multiresolution gray-scale and rotation invariant texture classification with local binary patterns
Ojala, T., Pietikainen, M., & Maenpaa, T. (2002) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2002.1017623 · Cited by 15,398 · cited 1×
“The authors of [8], in turn, proposed to use one-class approach us-ing local binary pattern [9] of linear frequency cepstral coef-ficients (LFCCs) followed by support vector machine (SVM) for voice conversion, speech synthesis and artificial signal detection [8].”
MethodFundamentals of speech recognition
Rabiner, L. R., & Juang, B.-H. F. (1993) · Cited by 7,680
Davis, S., & Mermelstein, P. (1980) · IEEE Transactions on Acoustics, Speech, and Signal Processing · doi:10.1109/tassp.1980.1163420 · Cited by 5,465 · cited 1×
“In MFCC, the filters are placed in mel scale, having denser spacing in t he low-frequency region [21].”
MethodPerceptual linear predictive (PLP) analysis of speech
Hermansky, H. (1990) · The Journal of the Acoustical Society of America · doi:10.1121/1.399423 · Cited by 2,536 · cited 1×
“Another all-pole representation of speech called perceptual linear prediction cepstral coefficients (PLPCC) is also computed by first performing a series of perceptual processing prior to LP analysis [24].”
MethodCepstral analysis technique for automatic speaker verification
Furui, S. (1981) · IEEE Transactions on Acoustics, Speech, and Signal Processing · doi:10.1109/tassp.1981.1163530 · Cited by 1,173
Continuous probabilistic transform for voice conversion
Stylianou, Y., Cappe, O., & Moulines, E. (1998) · IEEE Transactions on Speech and Audio Processing · doi:10.1109/89.661472 · Cited by 931 · cited 1×
“Synthetic speech signal created using different voice conv ersion (VC) and speech synthesis (SS) techniques can be used to poof biometric systems based on automatic speaker recognition t echnology [1, 2, 3, 4].”
MethodConstruction and evaluation of a robust multifeature speech/music discriminator
Scheirer, E., & Slaney, M. (2002) · 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing · doi:10.1109/icassp.1997.596192 · Cited by 873 · cited 1×
“Spectral flux based feature: Spectral flux measures the frame-by-frame change in the power spectrum [25].”
BackgroundCOVAREP — A collaborative voice analysis repository for speech technologies
Degottex, G., Kane, J., Drugman, T., Raitio, T., & Scherer, S. (2014) · 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2014.6853739 · Cited by 695 · cited 1×
“2, First 20 coefficients are retained after DCT [5] Short-term APGDF (60) LP Order=20 — phase features CosPhase (60) First 20 coefficients are retained after DCT [5] RPS (60) RPS computed with COVAREP tool [14], 20 filters in mel filter ba nk [6, 7]”
MethodSpoofing and countermeasures for speaker verification: A survey
Wu, Z., Evans, N., Kinnunen, T., Yamagishi, J., Alegre, F., & Li, H. (2015) · Speech Communication · doi:10.1016/j.specom.2014.10.005 · Cited by 601
Self-citationThe cepstrum: A guide to processing
Childers, D. G., Skinner, D. P., & Kemerait, R. C. (1977) · Proceedings of the IEEE · doi:10.1109/proc.1977.10747 · Cited by 600 · cited 1×
“Cepstrum: Cepstral coefficients (Cep) are computed from the power spectrum by applying discrete cosine transform (DCT) [17].”
Method
Cited by
Items 101–200 of 383
Cross-modal information fusion for voice spoofing detection
Xue, J., Zhou, H., Song, H., Wu, B., & Shi, L. (2023) · Speech Communication · doi:10.1016/j.specom.2023.01.001 · Cited by 34
Spoofing Detection on the ASVspoof2015 Challenge Corpus Employing Deep Neural Networks
Alam, J., Kenny, P. J., Gupta, V., & Stafylakis, T. (2016) · The Speaker and Language Recognition Workshop (Odyssey) · doi:10.21437/odyssey.2016-39 · Cited by 31
“The motivation for excluding the static feature and using only delta or delta + double delta features for this task is the findings of [27].”
BackgroundGeneralized Spoofing Detection Inspired from Audio Generation Artifacts
Gao, Y., Vuong, T., Elyasi, M., Bharaj, G., & Singh, R. (2021) · Interspeech 2021 · doi:10.21437/interspeech.2021-1705 · Cited by 25
A New Replay Attack Against Automatic Speaker Verification Systems
Yoon, S.-H., Koh, M.-S., Park, J., & Yu, H.-J. (2020) · IEEE Access · doi:10.1109/access.2020.2974290 · Cited by 34
“Regardless of the feature extraction method, we did not perform voice activity detection on all speech samples, because silent segments are also useful for spoofing detection [22].”
MethodDeep learning technology for face forgery detection: A survey
Ma, L., Yang, P., Xu, Y., Yang, Z., Li, P., & Huang, H. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.129055 · Cited by 41
Robust features fusion for text independent speaker verification enhancement in noisy environments
Mohammadi, M., & Mohammadi, H. R. S. (2017) · Iranian Conference on Electrical Engineering · doi:10.1109/iraniancee.2017.7985357 · Cited by 31
Two-level fusion-based acoustic scene classification
Waldekar, S., & Saha, G. (2020) · Applied Acoustics · doi:10.1016/j.apacoust.2020.107502 · Cited by 25
The biometric vox system for the ASVspoof 2021 challenge
Cáceres, J., Font, R., Grau, T., & Molina, J. (2021) · 2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge · doi:10.21437/asvspoof.2021-11 · Cited by 25
“…For this system, 30 static Constant-Q Cepstral Coefficients (CQCCs) [13] are computed with f min = 15, f max =8000 Hz and 96 bins per octave. tdnn-SCMC system : Previous work [14, 15] has shown that Spectral Centroid Magnitude Coefficients (SCMCs) can provide a robust front-end for spoofing detection.”
MethodNew Acoustic Features for Synthetic and Replay Spoofing Attack Detection
Wei, L., Long, Y., Wei, H., & Li, Y. (2022) · Symmetry · doi:10.3390/sym14020274 · Cited by 29
“The constant Q cepstral coefficients (CQCCs) [9] and linear frequency cepstral coefficients (LFCCs) [30] are two types of effective conventional acoustic features used for antispoofing speech detection in ASV tasks.”
BackgroundMethodInfluentialDuan, D., Sun, Z., Ni, T., Li, S. C., Jia, X., Xu, W., & Li, T. (2024) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3643832.3661860 · Cited by 42
Voice Conversion Using Speech-to-Speech Neuro-Style Transfer
AlBadawy, E. A., & Lyu, S. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-3056 · Cited by 24
“The AVSspoof 2019 Challenge [11] provides a Gaussian Mixture Model (GMM)-based model as their main classifier with linear frequency cepstral coefficients [21] (LFCC) features.”
Audio Deepfake Detection System with Neural Stitching for ADD 2022
Yan, R., Cheng, W. M., Zhou, S., Guo, T., Zou, W., & Li, X. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746820 · Cited by 38
Classification of audio scenes with novel features in a fused system framework
Waldekar, S., & Saha, G. (2018) · Digital Signal Processing · doi:10.1016/j.dsp.2017.12.012 · Cited by 24
Replay spoofing countermeasure using autoencoder and siamese networks on ASVspoof 2019 challenge
Adiban, M., Sameti, H., & Shehnepoor, S. (2020) · Computer Speech & Language · doi:10.1016/j.csl.2020.101105 · Cited by 29
Kinnunen, T., Lorenzo-Trueba, J., Yamagishi, J., Toda, T., Saito, D., Villavicencio, F., & Ling, Z. (2018) · The Speaker and Language Recognition Workshop (Odyssey 2018) · doi:10.21437/odyssey.2018-27 · Cited by 23
Self-citationPhysiological-physical feature fusion for automatic voice spoofing detection
Xue, J., & Zhou, H. (2022) · Frontiers of Computer Science · doi:10.1007/s11704-022-2121-6 · Cited by 28
“At present, the common methods are MFCC, LPC, LPCC, LFCC, etc [25].”
MethodNovel speech features for improved detection of spoofing attacks
Paul, D., Pal, M., & Saha, G. (2015) · IEEE India Conference · doi:10.1109/indicon.2015.7443805 · Cited by 24
“features for spoofing detection system was given in [21].”
BackgroundMethodSpoofing speech detection using temporal convolutional neural network
Tian, X., Xiao, X., Chng, E. S., & Li, H. (2016) · 2016 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA) · doi:10.1109/apsipa.2016.7820738 · Cited by 23
“cepstrum coefficient (MFCC) [14, 15], linear frequency cepstral coefficients (LFCC) [16], Subband centroid frequency coefficient (SCFC) [17] and so on.”
BackgroundA Deeper Look at Gaussian Mixture Model Based Anti-Spoofing Systems
Chettri, B., & Sturm, B. L. T. (2018) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2018.8461467 · Cited by 33
“We find the optimal number of components for each GMM as: 512 for MFCC, LFCC and CQCC; 128 for IMFCC and RFCC; and 256 for SCMC. Table 2 shows the results of six GMM systems on both the development and evaluation datasets.”
MethodInfluentialReplay Attack Detection Using Linear Prediction Analysis-Based Relative Phase Features
Phapatanaburi, K., Wang, L., Nakagawa, S., & Iwahashi, M. (2019) · IEEE Access · doi:10.1109/access.2019.2960369 · Cited by 25
“MFCC features [47] are extracted using a 20 ms frame length and 10 ms frameshift.”
MethodSinha, S., Dey, S., & Saha, G. (2023) · Computer Speech & Language · doi:10.1016/j.csl.2023.101599 · Cited by 33
Alam, J., & Kenny, P. J. (2017) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco.2017.8081177 · Cited by 22
“Besides CQCC features we also considered following features for comparison purpose - Mel frequency cepstral coefficients (MFCC), linear frequency cepstral coefficients (LFCC), linear predictive cepstral coefficients (LPCC), modified group delay [27]-based cepstral coefficients (MGDC), spectral flux cepstral coefficien…”
BackgroundMethodMulti-Task Learning Improves Synthetic Speech Detection
Mo, Y., & Wang, S. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746059 · Cited by 30
Generalized Voice Spoofing Detection via Integral Knowledge Amalgamation
Ren, Y., Peng, H., Li, L., Xue, X., Yang, L., & Yang, Y. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3288416 · Cited by 25
“Among them, Linear Frequency Cepstral coefficients (LFCC) [7] has good detection performance in ASVspoof 2019 Challenge.”
BackgroundInfluentialLi, J., Wang, H., He, P., Abdullahi, S. M., & Li, B. (2022) · Digital Signal Processing · doi:10.1016/j.dsp.2021.103256 · Cited by 20
Wang, B., Tang, Y., Wei, F., Ba, Z., & Ren, K. (2024) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2024.3492796 · Cited by 20
Multi-task Learning-Based Spoofing-Robust Automatic Speaker Verification System
Zhao, Y., Togneri, R. B., & Sreeram, V. (2022) · Circuits Systems and Signal Processing · doi:10.1007/s00034-022-01974-z · Cited by 26
“of filters = 20/frame size = 20ms/frame shift = 10ms/FFT bins = 512 [43]”
Mittal, A., & Dua, M. (2021) · Advances in intelligent systems and computing · doi:10.1007/978-981-15-8443-5_76 · Cited by 20
ADD 2023: the Second Audio Deepfake Detection Challenge
Yi, J., Tao, J., Fu, R., Yan, X., Wang, C., Wang, T., Zhang, C. Y., Zhang, X., Zhao, Y., Ren, Y., et al. (2023) · arXiv · doi:10.48550/arxiv.2305.13774 · Cited by 20
Potential of Speech-Pathological Features for Deepfake Speech Detection
Chaiwongyen, A., Duangpummet, S., Karnjana, J., Kongprawechnon, W., & Unoki, M. (2024) · IEEE Access · doi:10.1109/access.2024.3447582 · Cited by 28
“Numerous features have been used for detecting deepfake speech, including spectrograms, linear-frequency cepstral coefficients (LFCCs) [25], mel-frequency cepstral coefficients [26], constant-Q transform [27], and constant-Q cepstral coefficients [28].”
MethodOn the use of convolutional neural networks for speech presentation attack detection
Korshunov, P., Gonçalves, A. R., Violato, R. P. V., Simões, F. O., & Marcel, S. (2018) · IEEE International Conference on Identity, Security and Behavior Analysis (ISBA) · doi:10.1109/isba.2018.8311474 · Cited by 19
“For the baseline state of the art PAD system, we selected a system based on mel-scale frequency coefficients (MFCC) [2] and Gaussian mixture model (GMM)-based classifier, which was shown to perform well by [18, 10, 11] in single and cross-database scenarios.”
BackgroundMethodInfluentialPatil, A. T., Acharya, R., Patil, H. A., & Guido, R. C. (2021) · Computer Speech & Language · doi:10.1016/j.csl.2021.101281 · Cited by 20
Comparative Analysis of ASV Spoofing Countermeasures: Evaluating Res2Net-Based Approaches
Yang, M., Zheng, K., Wang, X., Sun, Y., & Chen, Z. (2023) · IEEE Signal Processing Letters · doi:10.1109/lsp.2023.3311367 · Cited by 20
“The LFCC is a 60-dimensional feature vector provided by the ASVspoof 2019 official [4].”
Lightweight Voice Spoofing Detection Using Improved One-Class Learning and Knowledge Distillation
Ren, Y., Peng, H., Li, L., & Yang, Y. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3321505 · Cited by 27
“…versatility and effectiveness of the proposed DOC-Softmax for different voice anti-spoofing systems, we adopt ResNet-CBAM [17], ResNet-18 [18], LCNN [21] and SE-Res2Net50 [22] as back-end classifiers, and LFCC [9] and CQT as front-end features to form different anti-spoofing systems for experiments.”
BackgroundMethodData selection for i-vector based automatic speaker verification anti-spoofing
Hanilçi, C. (2017) · Digital Signal Processing · doi:10.1016/j.dsp.2017.10.010 · Cited by 22
Self-citationRobustness of Speech Spoofing Detectors Against Adversarial Post-Processing of Voice Conversion
Ding, Y., Lin, H.-J., Liu, L., Ling, Z.-H., & Hu, Y. (2021) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2021.3124420 · Cited by 24
GANBA: Generative Adversarial Network for Biometric Anti-Spoofing
Gomez-Alanis, A., González, J. A., & Peinado, A. M. (2022) · Applied Sciences · doi:10.3390/app12031454 · Cited by 27
“features have been proposed for training these models, such as linear frequency cepstral coefficients (LFCC) [32], spectrograms [33], constant Q cepstral coefficients (CQCC) [34], and raw speech samples [35].”
BackgroundImproving anti-spoofing with octave spectrum and short-term spectral statistics information
Yang, J., & Das, R. K. (2019) · Applied Acoustics · doi:10.1016/j.apacoust.2019.107017 · Cited by 18
Muckenhirn, H., Magimai.-Doss, M., & Marcel, S. (2016) · International Conference of the Biometrics Special Interest Group (BIOSIG) · doi:10.1109/biosig.2016.7736933 · Cited by 19
“A performance comparison of 19 features on the ASVspoof database can be found in [20].”
BackgroundSignificance of Source–Filter Interaction for Classification of Natural vs. Spoofed Speech
Patel, T. B., & Patil, H. A. (2017) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2017.2682788 · Cited by 19
“It has been observed that the higher frequency regions of speech are essential for the SSD task [23].”
BackgroundMethodDeep domain adaptation for anti-spoofing in speaker verification systems
Himawan, I., Villavicencio, F., Sridharan, S., & Fookes, C. (2019) · Computer Speech & Language · doi:10.1016/j.csl.2019.05.007 · Cited by 21
Voice conversion spoofing detection by exploring artifacts estimates
Hemavathi, R., & Kumaraswamy, R. (2021) · Multimedia Tools and Applications · doi:10.1007/s11042-020-10212-0 · Cited by 24
Densely Connected Convolutional Network for Audio Spoofing Detection
Wang, Z., Cui, S., Kang, X., Sun, W., & Li, Z. (2020) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · Cited by 37
Monteiro, J., Alam, J., & Falk, T. H. (2020) · ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp40776.2020.9054558 · Cited by 17
“We thus follow the findings reported in [16] and use both linear frequency cesptral coefficients [17] (LFCC), which showed to be well suited for detecting LA attacks, and the product spectrum (ProdSpec) introduced in [18] and later used in [19] as a countermeasure for replay spoofing detection.”
MethodWani, T. M., Qadri, S. A. A., Wani, F. A., & Amerini, I. (2024) · Foundations and Trends® in Privacy and Security · doi:10.1561/3300000048 · Cited by 16
Complementary regional energy features for spoofed speech detection
Dişken, G. (2023) · Computer Speech & Language · doi:10.1016/j.csl.2023.101602 · Cited by 17
Audio Multi-View Spoofing Detection Framework Based on Audio-Text-Emotion Correlations
Wu, J., Yin, Q., Sheng, Z., Lu, W., Huang, J., & Li, B. (2024) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2024.3431888 · Cited by 24
“…a back-end machine learning classifier (such as a support vector machine or a Gaussian mixture model) or a DNN-based decoder have become routine solutions for audio spoofing detection [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34].”
BackgroundSpoofing countermeasure for fake speech detection using brute force features
Mirza, A. R., & Al‐Talabani, A. K. (2024) · Computer Speech & Language · doi:10.1016/j.csl.2024.101732 · Cited by 18
Audio Deep Fake Detection with Sonic Sleuth Model
Alshehri, A. S., Almalki, D., Alharbi, E., & Albaradei, S. (2024) · Computers · doi:10.3390/computers13100256 · Cited by 27
Demiroğlu, C., Büyük, O., Khodabakhsh, A., & Maia, R. (2017) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2017.2673807 · Cited by 17
“Comparison of a wide range of phase and magnitude spectrum features for SSDs is done in [30] and [31].”
BackgroundMonteiro, J., Alam, J., & Falk, T. H. (2019) · Machine learning for signal processing ... · doi:10.1109/mlsp.2019.8918703 · Cited by 22
“Specifically, we report results for models trained on top of the power spectrum (Spec), the product spectrum (ProdSpec) introduced in [10] and later used in [11] as a countermeasure for replay spoofing detection, linear frequency cesptral coefficients [12] (LFCC), and the constant Q cesptral coefficients (CQCC) [13].”
MethodMewada, H. K., Al‐Asad, J. F., Almalki, F. A., Khan, A. H., Almujally, N. A., El-Nakla, S., & Naith, Q. H. (2023) · Sensors · doi:10.3390/s23146637 · Cited by 20
“The authors in Reference [17] presented several spectral features to classify genuine audio from spoofed audio.”
BackgroundMethodVoice spoofing detection for multiclass attack classification using deep learning
Boyd, J., Fahim, M., & Olukoya, O. (2023) · Machine Learning with Applications · doi:10.1016/j.mlwa.2023.100503 · Cited by 29
On the Generalization of Fused Systems in Voice Presentation Attack Detection
Gonçalves, A. R., Violato, R. P. V., Korshunov, P., Marcel, S., & Simões, F. O. (2017) · International Conference of the Biometrics Special Interest Group (BIOSIG) · doi:10.23919/biosig.2017.8053516 · Cited by 13
“Despite the severity of the problem, researchers started to develop effective presentation attack detection mechanisms only in the last few years [1].”
BackgroundInfluentialMonteiro, J., & Alam, J. (2019) · IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU) · doi:10.1109/asru46091.2019.9003792 · Cited by 17
“• LFCC-FV/PLDA (single best): Selected single best system for LA task.”
MethodInfluentialCRIM's System Description for the ASVSpoof2021 Challenge
Kang, W. H., Alam, J., & Fathan, A. (2021) · doi:10.21437/asvspoof.2021-16 · Cited by 15
Feature Genuinization based Residual Squeeze-and-Excitation for Audio Anti-Spoofing in Sound AI
Ray, R., Karthik, S., Mathur, V., Kumar, P., Maragatham, G., Tiwari, S., & Shankarappa, R. T. (2021) · International Conference on Computing, Communication and Networking Technologies (ICCCNT) · doi:10.1109/icccnt51525.2021.9580127 · Cited by 15
Linear prediction residual features for automatic speaker verification anti-spoofing
Hanilçi, C. (2017) · Multimedia Tools and Applications · doi:10.1007/s11042-017-5181-0 · Cited by 15
“The pre-emphasis filter is used as a pre-processing step because according to previous results reported in [28] and our initial experiments on ASVspoof 2015 database, the high frequency components”
MethodResultSelf-citationSmartphone Multi-modal Biometric Authentication: Database and EvaluationPreprint
Ramachandra, R., Stokkenes, M., Mohammadi, A., Venkatesh, S. K., Raja, K. B., Wasnik, P., Poiret, E., Marcel, S., & Busch, C. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1912.02487 · Cited by 27
“b) Audio-visual attack instruments: We use three features: MFCC,SCFC,LFCC with two classifiers (SVM and GMM) [36] to develop voice PAD systems.”
Introduction to Voice Presentation Attack Detection and Recent Advances
Sahidullah, M., Delgado, H., Todisco, M., Nautsch, A., Wang, X., Kinnunen, T., Evans, N., Yamagishi, J., & Lee, K. A. (2023) · Advances in computer vision and pattern recognition · doi:10.1007/978-981-19-5288-3_13 · Cited by 22
Self-citationA Countermeasure Based on CQT Spectrogram for Deepfake Speech Detection
Ziabary, P. A., & Veisi, H. (2021) · International Conference on Signal Processing and Intelligent Systems (ICSPIS) · doi:10.1109/icspis54653.2021.9729387 · Cited by 23
“A comparison of different acoustic features for synthetic speech detection has been made in [25], using GMM and support vector machine for classification.”
MethodFeatures and classifiers for replay spoofing attack detection
Hanilçi, C. (2017) · Bursa Technical University Institutional Repository · Cited by 17
Self-citationCollaborative Watermarking for Adversarial Speech Synthesis
Juvela, L., & Wang, X. (2024) · ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp48485.2024.10448134 · Cited by 13
“Its front end extracts linear frequency cepstrum coefficients (LFCC) [25], which is similar to the Mel frequency cepstrum coefficients but uses filters placed in equal sizes on a linear scale.”
MethodMultiple Points Input For Convolutional Neural Networks in Replay Attack Detection
Yoon, S.-H., & Yu, H.-J. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9053303 · Cited by 15
“We do not perform voice activity detection [19] nor cepstral mean and variance normalization.”
MethodJoint operation of voice biometrics and presentation attack detection
Korshunov, P., & Marcel, S. (2016) · IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS) · doi:10.1109/btas.2016.7791179 · Cited by 14
“The benchmarking study on ‘logical access’ attacks [11] finds GMMs to be more successful compared to two-class SVM (combined with an LBP-based feature extraction from [1]) in detecting synthetic spoofing attacks.”
Two-Path GMM-ResNet and GMM-SENet for ASV Spoofing Detection
Lei, Z., Yan, H., Liu, C., Ma, M., & Yang, Y. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746163 · Cited by 19
“However, the ASV technologies are still vulnerable to some real-life spoofing attacks [2], such as mimicry [3], text-to-speech [4], voice conversion [5], and replay [6].”
BackgroundMethodCQT-Based Cepstral Features for Classification of Normal vs. Pathological Infant Cry
Patil, H. A., Kachhi, A., & Patil, A. T. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3325971 · Cited by 12
FairSSD: Understanding Bias in Synthetic Speech Detectors
Singh Yadav, A. K., Bhagtani, K., Salvi, D., Bestagini, P., & Delp, E. J. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00445 · Cited by 12
“000% 100.00% 0.000% 100.00% 0.000% 100.00% 0.000% 100.00% 0.000% 100.00% 0.000% LFCC-GMMs 𝐹𝑃𝑅 2 100.00% 0.”
MethodInfluentialThe DKU-LENOVO Systems for the INTERSPEECH 2019 Computational Paralinguistic Challenge
Wu, H., Wang, W., & Li, M. (2019) · Interspeech (USB) · doi:10.21437/interspeech.2019-1386 · Cited by 12
“And the results of the FV system are produced with Fisher Vectors encoding MFCC, LFCC, IMFCC, MGDF, and spectrum/cepstrum of PLP/RASTA-PLP [14].”
MethodInfluentialThe Sound of Silence: Efficiency of First Digit Features in Synthetic Audio Detection
Mari, D., Latora, F., & Milani, S. (2022) · IEEE International Workshop on Information Forensics and Security · doi:10.1109/wifs55849.2022.9975404 · Cited by 17
“This representation has already proved very effective in highlighting the more meaningful frequency elements in audio signals and in detecting forged waveforms [35].”
BackgroundInfluentialSailor, H. B., Kamble, M. R., & Patil, H. A. (2017) · Interspeech (USB) · doi:10.21437/interspeech.2017-1393 · Cited by 12
“The detailed analysis of various features for ASVspoof 2015 challenge database is presented in [7].”
BackgroundA Cross-Database Study of Voice Presentation Attack Detection
Korshunov, P., & Marcel, S. (2019) · Advances in computer vision and pattern recognition · doi:10.1007/978-3-319-92627-8_16 · Cited by 15
Firc, A., Malinka, K., & Hanáček, P. (2025) · Cybersecurity · doi:10.1186/s42400-024-00346-1 · Cited by 34
“Similarly, works by Li et al. (2022) and Sahidullah et al. (2015) examine the performance of various features that serve as input for deepfake speech detectors.”
Wang, Z., & Hansen, J. H. L. (2024) · IEEE Access · doi:10.1109/access.2024.3421281 · Cited by 19
“Such spoofed speech generated by different attacking algorithms contains artefacts, which reside in specific sub-bands or temporal segments [13], [14], [15], [16], [17], [18], [19].”
BackgroundEmoFake: An Initial Dataset for Emotion Fake Audio Detection
Zhao, Y. C., Yi, J., Tao, J., Wang, C., & Dong, Y. (2024) · Lecture notes in computer science · doi:10.1007/978-981-97-8367-0_25 · Cited by 30
Robust Audio Anti-Spoofing with Fusion-Reconstruction Learning on Multi-Order Spectrograms
Wen, P., Hu, K., Yue, W., Zhang, S., Zhou, W., & Wang, Z. (2023) · Interspeech (USB) · doi:10.21437/interspeech.2023-563 · Cited by 20
“Early studies primarily focused on de-vising hand-crafted features to anti-spoofing, such as Cochlear Filter Cepstral Coefficient Instantaneous Frequency (CFCCIF) [7], Linear Frequency Cepstral Coefficients (LFCC) [8] and Constant-Q Cepstral Coefficients (CQCC) [9].”
BackgroundAn Experimental Study on Audio Replay Attack Detection Using Deep Neural Networks
Bakar, B., & Hanilçi, C. (2018) · IEEE Spoken Language Technology Workshop (SLT) · doi:10.1109/slt.2018.8639511 · Cited by 12
“For example, it was found that the high-frequency region conveys more discriminative information for detecting SS and VC attacks for magnitude spectrum based features [14].”
BackgroundSelf-citationSpeech frame selection for spoofing detection with an application to partially spoofed audio-data
Kumar, A. K., Paul, D., Pal, M., Sahidullah, M., & Saha, G. (2021) · International Journal of Speech Technology · doi:10.1007/s10772-020-09785-w · Cited by 14
“Along with the popular speech features used in other speech applications, studies have been conducted for better representation of the speech signal for the spoofed speech detection task (Sahidullah et al. 2015; Paul et al. 2017a; Todisco et al. 2017; Patel and Patil 2017; Pal et al. 2018).”
Self-citationSpoof Detection using Voice Contribution on LFCC features and ResNet-34
Mon, K. Z., Galajit, K., Mawalim, C. O., Karnjana, J., Isshiki, T., & Aimmanee, P. (2023) · International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP) · doi:10.1109/isai-nlp60301.2023.10354625 · Cited by 11
“It has shown promising results for detecting physical access (PA) and logical access (LA) attacks [8] [9] [10] [11].”
BackgroundGeneralizing Speaker Verification for Spoof Awareness in the Embedding Space
Liu, X., Sahidullah, M., Lee, K. A., & Kinnunen, T. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3358056 · Cited by 22
“For instance, some CM modules usually require their specialized featureextractors(e.g. [21],[22])andclassifiers(e.g. [23]).”
BackgroundSelf-citationShahzad, K., Farhan, S., Haq, Y. U., Sana, R., & Pathan, M. S. (2025) · IEEE Access · doi:10.1109/access.2025.3544639 · Cited by 16
“A comparative analysis was conducted on 19 voice features to evaluate their performance on the synthetic speech ASV Spoof 2015 dataset [12].”
Voice-cloning artificial-intelligence speakers can also mimic human-specific vocal expression
Chen, W., & Jiang, X. (2026) · Current Psychology · doi:10.1007/s12144-026-09991-w · Cited by 17
“On the one hand, machine learning algorithms can reliably distinguish AI-generated speech from human speech (Sahidullah et al., 2015 ), based on systematic acoustic differences aris - ing from the inherently more linear generation process of synthetic voices compared to the nonlinear physiological…”
InfluentialLiu, L., & Yang, J. (2020) · IEEE Access · doi:10.1109/access.2020.3013066 · Cited by 10
“Many features such as linear frequency cepstral coefficient (LFCC), linear prediction cepstral coefficient (LPCC), perceptual PLPCC, mel frequency cepstral coeffi-cient (MFCC) and inverted MFCC have been investigated for the synthetic speech detection [15].”
BackgroundPatil, A. T., Patil, H. A., & Khoria, K. (2021) · Computer Speech & Language · doi:10.1016/j.csl.2021.101301 · Cited by 13
LARIHS ASSERT Reassessment for Logical Access ASVspoof 2021 Challenge
Benhafid, Z., Selouani, S., Yakoub, M. S., & Amrouche, A. (2021) · doi:10.21437/asvspoof.2021-15 · Cited by 10
“Moreover, most systems in this context demonstrate that linear frequency cepstral coefficients (LFCCs) [8] remain the best feature extraction front-end.”
BackgroundMethodMonteiro, J., Alam, J., & Falk, T. H. (2022) · Speech Communication · doi:10.1016/j.specom.2022.03.008 · Cited by 11
Saito, Y., Takamichi, S., & Saruwatari, H. (2019) · Computer Speech & Language · doi:10.1016/j.csl.2019.05.008 · Cited by 11
Contreras, R. C., Viana, M. S., Fonseca, E. S., dos Santos, F. L., Zanin, R. B., & Guido, R. C. (2023) · Sensors · doi:10.3390/s23115196 · Cited by 13
“• Cepstral features: For CC extraction techniques, the following methods will be considered: CQCC, MFCC, inverse MFCC (iMFCC), linear-frequency cepstral coefficients (LFCC) [90], gammatone-frequency cepstral coefficients (GFCCs) [91], bark-frequency cepstral coefficients (BFCCs) [92], LPCC [93], and normalized gammach…”
MethodInfluentialSubband fusion of complex spectrogram for fake speech detection
Fan, C., Xue, J., Dong, S., Ding, M., Yi, J., Li, J., & Lv, Z. (2023) · Speech Communication · doi:10.1016/j.specom.2023.102988 · Cited by 17
“Moreover, several studies (Patel and Patil, 2017; Sahidullah et al., 2015; Sriskandaraja et al., 2016; Witkowski et al., 2017; Garg et al., 2019; Lin et al., 2018; Yang et al., 2019; Soni et al., 2016; Chettri et al., 2020; Ling et al., 2021; Liu et al., 2021; Zhang et al., 2021)…”
BackgroundSDI: A tool for speech differentiation in user identification
Basit, M. A., Liu, C., & Zhao, E. (2022) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.122866 · Cited by 12
Machine Learning Mitigants for Speech Based Cyber Risk
Campi, M., Peters, G. W., Azzaoui, N., & Matsui, T. (2020) · IEEE Access · doi:10.1109/access.2021.3117080 · Cited by 11
Impact of Bandwidth and Channel Variation on Presentation Attack Detection for Speaker Verification
Delgado, H., Todisco, M., Evans, N., Sahidullah, M., Liu, W. M., Alegre, F., Kinnunen, T., & Fauve, B. (2017) · 2017 International Conference of the Biometrics Special Interest Group (BIOSIG) · doi:10.23919/biosig.2017.8053510 · Cited by 8
“While the original work [16] used 20 filters, use of a greater number was found to improve performance.”
BackgroundMethodInfluentialSelf-citationKhanjani, Z., Davis, L. S., Tuz, A., Nwosu, K., Mallinson, C., & Janeja, V. P. (2023) · IEEE International Conference on Intelligence and Security Informatics (ISI) · doi:10.1109/isi58743.2023.10297267 · Cited by 25
“Linear Frequency Cepstral Coefficient (LFCCs) [28] is the input of the Light Convolutional Neural Networks [20] as our baseline model.”
MethodASSD: An AI-Synthesized Speech Detection Scheme Using Whisper Feature and Types Classification
Liu, C., Xu, X., & Xiao, F. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2024.3520385 · Cited by 10
“Early feature-based synthesized speech detection methods relied on classical machine learning algorithms, such as the well-known Gaussian Mixture Model [27].”
Multi-Task Siamese Neural Network for Improving Replay Attack Detection
von Platen, P., Tao, F., & Tür, G. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-86 · Cited by 10
Robust Deepfake Audio Detection via Bi-Level Optimization
Li, M., Ahmadiadli, Y., & Zhang, X.-P. (2023) · 2023 IEEE 25th International Workshop on Multimedia Signal Processing (MMSP) · doi:10.1109/mmsp59012.2023.10337724 · Cited by 7
“…hand-crafted acoustic and perceptual features, such as Mel-frequency Cepstral Coefficients (MFCC) [8], [9], Constant Q Cepstral Coefficients (CQCC) [10], [11], and Perceptual Linear Predictive Coefficient (PLP) [12], have been widely used in the Deepfake detectors to extract audio representations.”
MethodTowards Attention-based Contrastive Learning for Audio Spoof Detection
Goel, C., Koppisetti, S., Colman, B., Shahriyari, A., & Bharaj, G. (2023) · INTERSPEECH 2023 · doi:10.21437/interspeech.2023-245 · Cited by 7
“Early works have focused on handcrafted features, such as cepstral coefficients [19], with classical methods such as Gaussian mixture models [20] (also see [8] and references therein).”
MethodHuang, L. F., Yang, J., Zhao, J., & Wu, C. (2025) · Expert Systems · doi:10.1111/exsy.70054 · Cited by 11
Deepfake-speech Detection with Pathological Features and Multilayer Perceptron Neural Network
Chaiwongyen, A., Duangpummet, S., Karnjana, J., Kongprawechnon, W., & Unoki, M. (2023) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc58517.2023.10317331 · Cited by 8
“Also, many speech features, including spec-trograms, linear-frequency cepstral coefficients (LFCCs) [14], mel-frequency cepstral coefficients [15], constant-Q transform [16], and constant-Q cepstral coefficients [17], have been used.”
MethodSynthetic Speech Detection Based on the Temporal Consistency of Speaker Features
Zhang, Y., Li, Z., Lu, J., Wang, W., & Zhang, P. (2024) · IEEE Signal Processing Letters · doi:10.1109/lsp.2024.3381890 · Cited by 13
Metadata
- Authors
- Sahidullah, M., Kinnunen, T., & Hanilçi, C.
- Affiliations
- University of Eastern Finland (FI)
- Venue
- Interspeech 2015
- Publisher
- ISCA
- Published
- 2015-09-06
- Pages
- 2087-2091
- ISSN
- 2308-457X (ISSN-L)
- OpenAlex
- W2295634712
- Cited by
- 382
- References
- 36
- Citations by year
- 2026: 23 · 2025: 41 · 2024: 42 · 2023: 50 · 2022: 38 · 2021: 51 · 2020: 35 · 2019: 41 · 2018: 15 · 2017: 29 · 2016: 15 · 2015: 2
- FWCI
- 11.46
- Type
- conference paper
- Language
- en
- Open access
- green
- erepo.uef.fi
- Licence
- other-oa
- Locations
- Interspeech (USB) (publishedVersion)
- UEF eRepo (University of Eastern Finland) (submittedVersion, other-oa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Physical Sciences › Computer Science › Signal Processing › Biometric Identification and Security
- Keywords
- synthetic speech detection, spoofing attacks, anti-spoofing, speaker verification, voice conversion, speech synthesis, feature extraction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.