上一条：Phone-aware multi-task learning and length expanding for short-duration language recognition

下一条：Extraction of noise-robust speaker embedding based on generative adversarial networks