A four-stage multimodal ensemble with transformer-based deep fusion, soft annotation targets, and meta-classifier soup achieves top rank in the Interspeech 2025 speech emotion recognition challenge.
Overview The novelty of our approach stems from three key directions: 1) architecture, 2) data utilization and 3) training recipe
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MEDUSA: A Multimodal Deep Fusion Multi-Stage Training Framework for Speech Emotion Recognition in Naturalistic Conditions
A four-stage multimodal ensemble with transformer-based deep fusion, soft annotation targets, and meta-classifier soup achieves top rank in the Interspeech 2025 speech emotion recognition challenge.