Recognition of Human Inner Emotion Based on Two-Stage FCA-ReliefF Feature Optimization

Currently, there is a growing interesting in emotion recognition. Representation of emotional states very challenging issue. Considering the calculation cost and generalization capability for practical application, series features which contain common time frequency domain are extracted from physiological signals to represent different states. To reduce feature dimensionality improve recognition accuracy, two-stage optimization method based on correlation analysis (FCA) ReliefF algorithm proposed select critical features. Firstly, FCA employed analyze redundancy between features, then adopted categories, optimal subset obtained using FCA-ReliefF method. Support vector machine as classifier evaluate classification performance this investigation. The effectiveness validated by testing two publicly available multimodal datasets, Augsburg Biosignal Toolbox (AuBT) Database Emotion Analysis Using Physiological Signals (DEAP). Compared with recent similar reported studies, developed research stable competitive, its accuracy reaches 98.40% 92.34%