Continuous Learning Based Novelty Aware Emotion Recognition System
This addresses the issue of model robustness in emotion recognition for real-world applications, but it appears incremental as it applies existing continuous learning methods to this domain.
The authors tackled the problem of emotion recognition models failing on novel data distributions by proposing a continuous learning approach, resulting in a system that adapts to unexpected samples.
Current works in human emotion recognition follow the traditional closed learning approach governed by rigid rules without any consideration of novelty. Classification models are trained on some collected datasets and expected to have the same data distribution in the real-world deployment. Due to the fluid and constantly changing nature of the world we live in, it is possible to have unexpected and novel sample distribution which can lead the model to fail. Hence, in this work, we propose a continuous learning based approach to deal with novelty in the automatic emotion recognition task.