Prechter Program faculty and affiliate researchers recently published their paper, "Improving Cross-Corpus Speech Emotion Recognition with Adversarial Discriminative Domain Generalization (ADDoG)." It has been selected at the runner-up for the 2021 Best Paper Award from IEEE Transactions on Affective Computing. Dr. Emily Mower Provost, co-author of the paper, summarizes the findings:
When our environment changes, the sound of our emotion expressions change as well. However, as humans, we know that this does not mean that in a new environment we lose the ability to recognize the emotion expressions of others. Yet, this is often the case for our automatic classifiers. In this paper, our goal is to create methods that identify commonalities between emotion expressions in different environments and to use these commonalities to recognize emotion in a more generalizable manner. We do this by creating a new “meet in the middle” approach that learn to identify commonalities between emotion expressions from different datasets. We referred to this method as Adversarial Discriminative Domain Generalization (ADDoG). It iteratively moves representations learned for each dataset (high-dimensional descriptions of data) closer to one another, improving cross-dataset generalization and significantly improving cross-dataset performance.