Efficient Speaker Verification System with Spoofing Attack
Malathi Sharavanan · 2014
In the existing system the state-of-the-art speaker verification system cannot distinguish the natural speech and converted speech. But in this project an attempt will be made to build an efficient speaker verification system to detect natural speech and synthesized speech. In this paper, we present new results evaluating the current state-of-the-art speaker verification system, Gaussian mixture model supervector with joint factor analysis (GMM-ZJFA) system, against spoofing attacks. In this the spoofing attacks are simulated by Gaussian mixture model based voice conversion technique. The results show that GMM-based conversion method which increases the false acceptance rate (FAR) from 3.24% to 17.33%. This suggests that GMM-JFA system is less vulnerable towards GMM-based conversion. The software used over here is matlab.