Spoofing-Robust Speaker Verification Based on Time-Domain Embedding

Abstract

Spoofing-robust speaker verification technology serves to safeguard voice-based authentication systems from fraudulent attempts. Such a system should be capable of detecting spoofed voice segments and verifying voice segments identified as genuine as originating from the real speaker. This research employs an understandable and explainable embedding based on the probability mass function of waveform amplitudes in the time domain. The results demonstrate that the performance of the countermeasure (CM) system is enhanced when it is gender dependent. The ASVspoof2019 challenge, logical access (LA) database was employed for evaluation purposes. The CM system demonstrated an equal error rate (EER) of 9.2% on the evaluation set for the male gender, with an EER of 10.1% for the female gender. In contrast, a gender-independent CM system exhibited an EER of 10.2%. The system’s performance, as quantified by the detection cost function for tandem assessment (t-DCF), is 0.262 for the gender-dependent system and 0.328 for the gender-independent system.

Tasks

Reproductions