Self-Supervised Frameworks for Speaker Verification via Bootstrapped Positive Sampling
This work addresses the problem of closing the performance gap with supervised systems in speaker verification for applications like biometrics, though it is incremental as it builds on existing SSL frameworks.
The paper tackles the limitation of similar channel characteristics in anchor-positive pairs for self-supervised speaker verification by introducing Self-Supervised Positive Sampling (SSPS), a bootstrapped technique that samples diverse positives to reduce channel information, achieving up to 58% relative reduction in EER and EERs as low as 2.53% on VoxCeleb benchmarks.
Recent developments in Self-Supervised Learning (SSL) have demonstrated significant potential for Speaker Verification (SV), but closing the performance gap with supervised systems remains an ongoing challenge. SSL frameworks rely on anchor-positive pairs, constructed from segments of the same audio utterance. Hence, positives have channel characteristics similar to those of their corresponding anchors, even with extensive data-augmentation. Therefore, this positive sampling strategy is a fundamental limitation as it encodes too much information regarding the recording source in the learned representations. This article introduces Self-Supervised Positive Sampling (SSPS), a bootstrapped technique for sampling appropriate and diverse positives in SSL frameworks for SV. SSPS samples positives close to their anchor in the representation space, assuming that these pseudo-positives belong to the same speaker identity but correspond to different recording conditions. This method consistently demonstrates improvements in SV performance on VoxCeleb benchmarks when applied to major SSL frameworks, including SimCLR, SwAV, VICReg, and DINO. Using SSPS, SimCLR and DINO achieve 2.57% and 2.53% EER on VoxCeleb1-O, respectively. SimCLR yields a 58% relative reduction in EER, getting comparable performance to DINO with a simpler training framework. Furthermore, SSPS lowers intra-class variance and reduces channel information in speaker representations while exhibiting greater robustness without data-augmentation.