Handcrafted features for anti spoofing

Abstract

Amongst various biometrics, voice is the most natural and convenient way of the newlinecommunication for human-machine interaction. To that effect, the use of Automatic newlineSpeaker Verification (ASV) for authentication is increasing in various sensitive newlineapplications, which create a chance for fraudulent attack as attackers can newlinebreach the authentication by using various spoofing attacks. To alleviate this issue, newlinewe can either develop an ASV system, which is inherently protected from newlinethe spoofing attacks or develop a separate countermeasure (CM) system that can newlineassist the ASV system in tandem against the spoofing attacks. The earlier approaches newlinehave trade-off between performance of the ASV system and robustness newlineagainst spoofing attacks. Hence, it would be advantageous to implement newlinethe separate Spoof Speech Detection (SSD) system, and hence majority research newlineattempts are focusing upon the later approach. To that effect, various international newlinechallenge campaigns were organized during INTERSPEECH conferences, newlinesuch as ASVSpoof 2015, ASVSpoof 2017, and ASVSpoof 2019, which provides newlinestandard datasets, protocol, and evaluation metrics. This thesis focuses on developing newlinethe handcrafted feature sets for CM systems against the spoofing attacks, newlinenamely, Speech Synthesis (SS), Voice Conversion (VC), and replay. These feature newlinesets are either developed by applying the subband filtering on the speech signals newlineor derived from the spectrogram representations. newlineIn this thesis work, various subband filtering-based feature sets are developed, newlinenamely, Enhanced Teager Energy-Based Cepstral Coefficients (ETECC), Cross- newlineTeager Energy Cepstral Coefficients (CTECC), and Energy Separation Algorithmbased newlineInstantaneous Frequency estimation for Cochlear Cepstral Features (CFCCIFESA). newlineThese feature sets are either modification in Teager Energy Operator (TEO)- newlinebased representations or utilization of Energy Separation Algorithm (ESA) for Instantaneous newlineFrequency (IF) estimation. The ETECC feature set is developed by newlineaccurately estimating the energies in high...