Support Vector Machine with Tunicate Swarm Optimization Algorithm for Emotion Recognition in Human-Robot Interaction

The rapid development of computer programs for the automatic classification of human emotions in recent years has drawn the attention of researchers. However, existing techniques have not addressed context-information included in facial expressions properly. In this research, a Tunicate Swarm Optimization Algorithm with Support Vector Machine (TSOA-SVM) to tackle the issue and enhance performance in emotion recognition. Then, ORL dataset was is used to the recommended approach to collect the data, then it was image-scaled and enhanced as part of the preparation process. Moreover, characteristics from previously processed images are extracted using Wavelet Transform and Entropy characteristics. After that, the classification stage is used to these derived traits to ascertain whether or not they include emotions. The proposed technique achieved high accuracy of 99.42%, specificity of 99.54%, and sensitivity of 99.35% in distinguishing the emotions when compared to other existing methods such as Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbour.