Speaker Verification
Speaker verification (SV) aims to automatically authenticate a person's identity based on their voice, focusing on creating robust and accurate systems. Current research emphasizes improving the discriminative power of speaker embeddings through techniques like contrastive learning, disentangling confounding factors such as age and channel variations, and leveraging powerful pre-trained models such as WavLM and Whisper. These advancements are crucial for enhancing security in various applications, from access control to forensic investigations, and are driving ongoing efforts to improve robustness against spoofing attacks and noisy conditions.
Papers
Model Compression for DNN-based Speaker Verification Using Weight Quantization
Jingyu Li, Wei Liu, Zhaoyang Zhang, Jiong Wang, Tan Lee
Convolution-Based Channel-Frequency Attention for Text-Independent Speaker Verification
Jingyu Li, Yusheng Tian, Tan Lee
Combining Automatic Speaker Verification and Prosody Analysis for Synthetic Speech Detection
Luigi Attorresi, Davide Salvi, Clara Borrelli, Paolo Bestagini, Stefano Tubaro
Wespeaker: A Research and Production oriented Speaker Embedding Learning Toolkit
Hongji Wang, Chengdong Liang, Shuai Wang, Zhengyang Chen, Binbin Zhang, Xu Xiang, Yanlei Deng, Yanmin Qian