Coming soon...
Uncertainty is Contagious: A novel generative modeling approach using soft-labels to model prediction uncertainties for speech emotion recognition
-
This study presents a generative modeling approach using VAEs to train emotion recognition models on soft-labels (true annotator distributions) to learn from the intricate confusion between labelers.
-
Presents an interest finding about using uncertainties to do transfer learning (more about this soon…)
Representation learning for affective speech signal analysis and processing
-
This is a tutorial paper on the speech emotion recognition solutions in our daily life that are addressed across three core dimensions (robustness, generalization and usability) using deep networks for affective speech modeling.
-
Presents different state of the art deep learning techniques with a focus is to learn speech representations that can be robust against settings of signal acquisition and the nature of emotion manifestation.
Uncertainty Estimation in Speech Emotion Recognition
-
Using Bayesian learning frameworks to model uncertainty in deep neural networks for speech emotion recognition.
-
Demonstrates the use of uncertainties for both emotion classification and emotional attribute regression scenarios and its application areas with use cases in various machine learning algorithms such as curriculum learning, co-training, active learning and multi-view training.
An Interpretable Deep Mutual Information Curriculum Metric for Robust and Generalized Speech Emotion Recognition
- In preparation…