Search papers, labs, and topics across Lattice.
Computer Science, Duke-Kunshan University, Kunshan, Suzhou 215316, China
3
1
5
10
The optimal spectrogram configuration for audio and speech analysis hinges on a nuanced interplay between front-end feature representation and back-end classifier architecture, varying significantly across tasks.
Adversarial training and synthetic data can significantly boost multilingual speaker verification performance, even with limited training data.
Forget hand-crafted curricula: TSE-Datamap leverages training dynamics to automatically surface optimal learning schedules for target speaker extraction.
