Affective Vocal Bursts (A-VB) Workshop and Competition
Affective Vocal Bursts (A-VB) Workshop and Competition
Introduction
The Affective Vocal Bursts (A-VB) Workshop and Competition (A-VB) was organized by Hume AI, a research lab and technology company. The workshop explores an under-studied markers for emotion recognition, that is the human’s non-verbal sound.
I (SclabCNU Team) paticipated in all 4 tasks of the competition and achieve top 2 in A-VB Culture task. The problem is designing a machine learning model to estimate the human behavior in different spaces. The detailed solution is described in my paper Fine-tuning Wav2vec for Vocal-burst Emotion Recognition. The paper is in proceeding of A-VB Workshop 2022.
I utilized the pre-trained Wav2Vec 2.0 provided by Pytorch audio library hub here
Model Architecture
Results
The best performances on validation set of the HUME-VB dataset are listed below:
| Task | Two | High | Culture | Type |
|---|---|---|---|---|
| Performance | 61.94 | 65.41 | 53.39 | 49.89 |