Date: 20th July 2023, 11:00 AM

 

Location: tba

 

Speaker: Dr. Xinzhou Xu

 

Abstract: This talk will briefly show our works in recent years on inter-emotion transfer in speech, which implicitly employ inter-modality transfer for learning low-resource emotional states in speech. Within the works in this talk, we aim to answered four questions: 1) Is antonomous zero-shot learning available in Speech Emotion Recognition (SER)? 2) Is conventional zero-shot learning effective in zero-shot SER tasks? 3) Which type of descriptors (Auditory Affective Descriptors; AADs) can better describe affective perception for spoken signals? 4) Can we achieve better performance for zero-shot SER tasks through reducing the gap between the descriptors and speech signals? Then, this talk will also present an overview for our furture possible works in relation to inter-emotion transfer in speech.