Automatic Speech Verification Spoofing Detection

View on arXiv ← Back to list

Authors: Shentong Mo, Haofan Wang, Pinxu Ren, Ta-Chung Chi

Published: 2020-12-15 05:18:09+00:00

AI Summary

This research paper investigates automatic speech verification spoofing detection using traditional machine learning models. The authors explore different audio features (MFCC and CQCC) and classifiers (SVM and GMM) to identify spoofed speech, evaluating performance using EER and t-DCF.

Abstract

Automatic speech verification (ASV) is the technology to determine the identity of a person based on their voice. While being convenient for identity verification, we should aim for the highest system security standard given that it is the safeguard of valuable digital assets. Bearing this in mind, we follow the setup in ASVSpoof 2019 competition to develop potential countermeasures that are robust and efficient. Two metrics, EER and t-DCF, will be used for system evaluation.

Key findings

MFCC features generally outperformed CQCC features in their experiments. SVM classifiers showed better accuracy than GMM classifiers, although GMM had lower EER. AdaBoost ensembling did not improve performance.

Approach

The approach uses a multi-model ensemble method. It extracts MFCC and CQCC features from audio samples and then employs SVM and GMM classifiers for spoof detection. The authors also investigate the use of AdaBoost for ensemble learning.

Datasets

ASVSpoof 2019 dataset, based on the VCTK corpus.

Model(s)

Support Vector Machine (SVM) and Gaussian Mixture Model (GMM). AdaBoost was used for ensembling SVM models.

Author countries

USA

← Previous