Key challenges in developing automatic emotion recognitionsystems include scarcity of labeled data and lack of gold-standard references . Even for the cues that are labeled as the same emotion category, thevariability of associated expressions can be high depending on the elicitationcontext . Ouremotion recognition model combines the gradient reversal technique with anentropy loss function as well as the softlabel loss . The experiment resultsshow that domain transfer learning methods can be employed to alleviate the mismatch between different elicitation approaches. Our work provides new insights into emotion data collection, particularly the impact of itselicitation strategies, and the importance of domain adaptation in emotionrecognition aiming for generalized systems. Our research provides newinsights into emotion .

Author(s) : Haoqi Li, Yelin Kim, Cheng-Hao Kuo, Shrikanth Narayanan

Links : PDF - Abstract

Code :
Coursera

Keywords : emotion - domain - loss - data - adaptation -

Leave a Reply

Your email address will not be published. Required fields are marked *