SSPS: Self-Supervised Positive Sampling for Robust Self-Supervised Speaker Verification

2025/05/20 by Lepage, Theo, Dehak, Reda · 1 citation
#Artificial Intelligence (cs.AI) #Audio and Speech Processing (eess.AS) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Sound (cs.SD) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2505.14561

Abstract

Self-Supervised Learning (SSL) has led to considerable progress in Speaker Verification (SV). The standard framework uses same-utterance positive sampling and data-augmentation to generate anchor-positive pairs of the same speaker. This is a major limitation, as this strategy primarily encodes channel information from the recording condition, shared by the anchor and positive. We propose a new positive sampling technique to address this bottleneck: Self-Supervised Positive Sampling (SSPS). For a given anchor, SSPS aims to find an appropriate positive, i.e., of the same speaker identity but a different recording condition, in the latent space using clustering assignments and a memory queue of positive embeddings. SSPS improves SV performance for both SimCLR and DINO, reaching 2.57% and 2.53% EER, outperforming SOTA SSL methods on VoxCeleb1-O. In particular, SimCLR-SSPS achieves a 58% EER reduction by lowering intra-speaker variance, providing comparable performance to DINO-SSPS.

Citations

• Self-Supervised Frameworks for Speaker Verification via Bootstrapped Positive Sampling
• Additive Margin in Contrastive Self-Supervised Frameworks to Learn Discriminative Speaker Representations
• GPS-SSL: Guided Positive Sampling to Inject Prior Into Self-Supervised Learning
• Self-Supervised Learning with Cluster-Aware-DINO for High-Performance Robust Speaker Verification
• Pushing the limits of self-supervised speaker verification using regularized distillation framework
• Self-Supervised Training of Speaker Encoder with Multi-Modal Diverse Positive Pairs
• Self-Supervised Speaker Verification Using Dynamic Loss-Gate and Label Correction
• Label-Efficient Self-Supervised Speaker Verification With Information Maximization and Contrastive Learning
• Curriculum learning for self-supervised speaker verification
• Self-Supervised Speaker Verification with Simple Siamese Network and Self-Supervised Regularization
• Emerging Properties in Self-Supervised Vision Transformers
• With a Little Help from My Friends: Nearest-Neighbor Contrastive Learning of Visual Representations
• Self-supervised Text-independent Speaker Verification using Prototypical Momentum Contrastive Learning
• Unsupervised Representation Learning for Speaker Recognition via Contrastive Equilibrium Learning
• Augmentation adversarial training for self-supervised speaker recognition
• A Simple Framework for Contrastive Learning of Visual Representations
• Momentum Contrast for Unsupervised Visual Representation Learning
• Momentum Contrast for Unsupervised Visual Representation Learning
• MUSAN: A Music, Speech, and Noise Corpus

Cited by

• MT-HuBERT: Self-Supervised Mix-Training for Few-Shot Keyword Spotting in Mixed Speech

Related

• Self-Supervised Frameworks for Speaker Verification via Bootstrapped Positive Sampling
• Self-Supervised Training of Speaker Encoder with Multi-Modal Diverse Positive Pairs
• Towards Supervised Performance on Speaker Verification with Self-Supervised Learning by Leveraging Large-Scale ASR Models
• Self-Distillation Prototypes Network: Learning Robust Speaker Representations without Supervision
• Why does Self-Supervised Learning for Speech Recognition Benefit Speaker Recognition?