Feedback Mechanism for Customer Care Service via Speech Emotion Recognition

Abstract

Speech Emotion Recognition is not a very active field of research. There are no standard models available. In this paper, we have used explicit feature engineering process instead of using a Neural Network model for feature extraction. Specific data augmentations are applied to generalize the data to a greater extent. A simple and more controllable yet, effective and efficient classifier is created with basic layers. The model contains just over 600K parameters. The accuracy achieved on unseen test data is 84%.