4 papers
Agentic-DuplexGen: Decoupling Content, Timing, and Acoustics for Synthetic Dialogue Speech
Pengcheng Wang, Sheng Li, Jiyi Li +1
Synthetic conversational speech has become an important resource for developing and evaluating conversational speech systems. However, existing dialogue synthesis pipelines typical…
FedCPC: An Effective Federated Contrastive Learning Method for Privacy Preserving Early-Stage Alzheimer's Speech Detection
Wenqing Wei, Zhengdong Yang, Yuan Gao +4
The early-stage Alzheimer's disease (AD) detection has been considered an important field of medical studies. Like traditional machine learning methods, speech-based automatic dete…
GhostVec: A New Threat to Speaker Privacy of End-to-End Speech Recognition System
Xiaojiao Chen, Sheng Li, Jiyi Li +3
Speaker adaptation systems face privacy concerns, for such systems are trained on private datasets and often overfitting. This paper demonstrates that an attacker can extract speak…
Reprogramming Self-supervised Learning-based Speech Representations for Speaker Anonymization
Xiaojiao Chen, Sheng Li, Jiyi Li +3
Current speaker anonymization methods, especially with self-supervised learning (SSL) models, require massive computational resources when hiding speaker identity. This paper propo…