Speech Representation Learning Revisited: The Necessity of Separate Learnable Parameters and Robust Data Augmentation

Open in new window