A Study of Using Cepstrogram for Countermeasure Against Replay Attacks

View on arXiv ← Back to list

Authors: Shih-Kuang Lee, Yu Tsao, Hsin-Min Wang

Published: 2022-04-09 00:18:53+00:00

AI Summary

This research demonstrates the effectiveness of cepstrograms as a countermeasure against replay attacks in automatic speaker verification. Experiments on the ASVspoof 2019 physical access database show that cepstrogram-based systems outperform other state-of-the-art methods, achieving the best results in both single and fusion systems.

Abstract

This study investigated the cepstrogram properties and demonstrated their effectiveness as powerful countermeasures against replay attacks. A cepstrum analysis of replay attacks suggests that crucial information for anti-spoofing against replay attacks may be retained in the cepstrogram. When building countermeasures against replay attacks, experiments on the ASVspoof 2019 physical access database demonstrate that the cepstrogram is more effective than other features in both single and fusion systems. Our LCNN-based single and fusion systems with the cepstrogram feature outperformed the corresponding LCNN-based systems without the cepstrogram feature and several state-of-the-art single and fusion systems in the literature.

Key findings

Cepstrograms proved more effective than other features (like LFCC, CQTgram, spectrogram) for replay attack detection. LCNN systems incorporating cepstrograms achieved state-of-the-art performance on the ASVspoof 2019 PA dataset, both individually and in a fusion setting. The cepstrogram's ability to highlight harmonic peaks caused by echoes in replayed speech is identified as the key to its effectiveness.

Approach

The researchers investigated the use of cepstrograms as features for detecting replay attacks. They integrated cepstrograms into a light convolutional neural network (LCNN)-based system, both as a single feature and in fusion with other features, for speaker verification.

Datasets

ASVspoof 2019 physical access database

Model(s)

Light Convolutional Neural Network (LCNN)

Author countries

Taiwan

← Previous