STC Anti-spoofing Systems for the ASVspoof 2015 Challenge

View on arXiv ← Back to list

Authors: Sergey Novoselov, Alexandr Kozlov, Galina Lavrentyeva, Konstantin Simonchik, Vadim Shchemelinin

Published: 2015-07-29 09:22:58+00:00

AI Summary

This paper details the Speech Technology Center's (STC) submissions to the ASVspoof 2015 challenge, focusing on exploring various acoustic feature spaces (MFCC, phase spectrum, wavelet transform) for robust spoofing detection. They employed TV-JFA for probability modeling and compared SVM and DBN classifiers.

Abstract

This paper presents the Speech Technology Center (STC) systems submitted to Automatic Speaker Verification Spoofing and Countermeasures (ASVspoof) Challenge 2015. In this work we investigate different acoustic feature spaces to determine reliable and robust countermeasures against spoofing attacks. In addition to the commonly used front-end MFCC features we explored features derived from phase spectrum and features based on applying the multiresolution wavelet transform. Similar to state-of-the-art ASV systems, we used the standard TV-JFA approach for probability modelling in spoofing detection systems. Experiments performed on the development and evaluation datasets of the Challenge demonstrate that the use of phase-related and wavelet-based features provides a substantial input into the efficiency of the resulting STC systems. In our research we also focused on the comparison of the linear (SVM) and nonlinear (DBN) classifiers.

Key findings
Wavelet-based features significantly outperformed MFCC features, achieving a 0.05% EER on the development dataset for known attacks. The best system (Primary system) achieved 1.965% EER on the evaluation dataset, but struggled with unknown attacks (3.92% EER), highlighting the need for more robust countermeasures for unseen attacks. SVM classifiers generally outperformed DBN classifiers.
Approach
The authors investigated different acoustic features (MFCC, phase spectrum features, wavelet-based features) for spoofing detection. They used a TV-JFA approach for probability modeling and compared linear (SVM) and non-linear (DBN) classifiers for classification. Feature fusion was also explored.
Datasets
ASVspoof 2015 Challenge development and evaluation datasets. The datasets contained 5 known spoofing attacks (S1-S5) and 5 unknown attacks (S6-S10).
Model(s)
TV-JFA (Total Variability Joint Factor Analysis) for probability modeling; SVM (Support Vector Machine) and DBN (Deep Belief Network) classifiers.
Author countries
Russia
← Previous