Vesper: A Compact and Effective Pretrained Model for Speech Emotion Recognition | IEEE Journals & Magazine | IEEE Xplore