Integration of CQCC and MFCC based Features for Replay Attack Detection

Abstract

This paper evaluates the performance of integration of CQCC and MFCC based features for automatic speaker verification (ASV) system. The detection of replay attack is challenging. For the detection of spoofing attacks, it is important to focus on front-end processing i.e., feature extraction. This paper discusses feature extraction techniques, LPC, MFCC and CQCC. The performance of baseline $\text{CQCC}+\text{GMM},\ \text{LPC}+\text{GMM}$, and $\text{MFCC}+\text{GMM}$ is evaluated on ASVspoof 2017 version 2 dataset. Further, integration of CQCC and MFCC showed improved performance resulting in EER 10.18%.