Using nonstandard SVM for combination of speaker verification and verbal information verification in speaker authentication system

Biometrics offers more security and convenience than traditional methods of identification. Recently, DNN has become a means of a more reliable and efficient authentication scheme. In this work, we compare two modern teaching methods: these two methods are methods based on the Gaussian mixture model (GMM) (denoted by the GMM i-vector) and methods based on deep neural networks (DNN) (denoted as the i-vector DNN). The results show that the DNN system with an i-vector is superior to the GMM system with an i-vector for various durations (from full length to 5s). DNNs have proven to be the most effective features for text-independent speaker verification in recent studies. In this paper, a new scheme is proposed that allows using DNN when checking text using hints in a simple and effective way. Experiments show that the proposed scheme reduces EER by 24.32% compared with the modern method and is evaluated for its reliability using noisy data, as well as data collected in real conditions. In addition, it is shown that the use of DNN instead of GMM for universal background modeling leads to a decrease in EER by 15.7%.

Download Full-text

Spoofing Detection and Countermeasure in Automatic Speaker Verification System using Dynamic Features

International Journal of Recent Technology and Engineering - 2 ◽

10.35940/ijrte.e6582.018520 ◽

2020 ◽

Vol 8 (5) ◽

pp. 3676-3680

Keyword(s):

Feature Extraction ◽

Speaker Verification ◽

Second Step ◽

Speech Sample ◽

Extraction Techniques ◽

Dynamic Features ◽

Universal Background Model ◽

Spoofing Detection ◽

Authentication System ◽

The Given

This present paper aims to extract robust dynamic features used to spoofing detection and countermeasure in ASV system. ASV is a biometric person authentication system. Researchers are aiming to develop spoofing detection and countermeasure techniques to protect this system against different spoofing attacks. For this, replayed attack is considered, because of very common accessibility of recording devices. In replay spoofing, the speech utterances of target (genuine) speakers are recorded and played against ASV system for gaining access unauthorizedly. For this purpose, as a first step, different dynamic features will be extracted for each speech sample. For feature extraction MFCC, LFCC, and MGDCC feature extraction techniques are used. As a second step, a classifier is used to classify whether the given speech sample is genuine or not. As a classifier, GMM and universal background model is used. In this present work, GMM based ASV system and Countermeasure systems using different feature extraction techniques are developed, and the performance of the methods is evaluated using EER and t- DCF. Basing on the performance values, the best feature extraction technique is selected.

Download Full-text