Paper proposes zero-shot cross-lingual speech emotion recognition model

Arxiv CS.SD (Sound/Audio ML)

arXiv:2606.06200 introduces a method for zero-shot cross-lingual speech emotion recognition (SER) that learns emotion-discriminative representations to handle distribution mismatches across languages. The model is trained only on source-language data and aims to generalize to target languages without emotion annotations. Authors include Jinyi Mi, Ding Ma, and Tomoki Toda.

speech-emotion-recognition

Arxiv CS.SD (Sound/Audio ML)·Jinyi MiDing MaTomoki Toda

6 days ago