Speech spoofing detection using SVM and ELM technique with acoustic features

Abstract

Now-a-days, the automatic speaker verification (ASV) systems are weak against attacks specially the voice conversion attacks and the speech synthesis attacks. To improve the robustness of the ASV systems, an anti-spoofing approach are developped to detect the spoofed speech from human speech. In this study, we focus on considering some acoustic features were proposed to differenciate spoofed speech from humain speech. We have used the proposed features with data from ASVspoof 2015 corpora. For the classification, we use Extreme learning machine (ELM) and Support Vector Machines (SVM) to obtain features and classified them to genuine or spoofed.