Multitask vocal burst modeling with ResNets and pre-trained paralinguistic Conformers
This work addresses the problem of modeling vocal bursts for tasks like emotion recognition, but it is incremental as it applies existing methods to a new competition dataset.
The authors tackled the ExVo-MultiTask competition by applying image classification models to mel-spectrograms of vocal bursts, achieving a 21.24% improvement over the baseline in task metrics, and found that single-task models outperformed multitask approaches.
This technical report presents the modeling approaches used in our submission to the ICML Expressive Vocalizations Workshop & Competition multitask track (ExVo-MultiTask). We first applied image classification models of various sizes on mel-spectrogram representations of the vocal bursts, as is standard in sound event detection literature. Results from these models show an increase of 21.24% over the baseline system with respect to the harmonic mean of the task metrics, and comprise our team's main submission to the MultiTask track. We then sought to characterize the headroom in the MultiTask track by applying a large pre-trained Conformer model that previously achieved state-of-the-art results on paralinguistic tasks like speech emotion recognition and mask detection. We additionally investigated the relationship between the sub-tasks of emotional expression, country of origin, and age prediction, and discovered that the best performing models are trained as single-task models, questioning whether the problem truly benefits from a multitask setting.