EmotionalGAN: Generating ECG to Enhance Emotion State Classification

Over the past few years, Generative Adversarial Networks (GANs) have been receiving attention from image and time series domain. In this work, we propose a novel sequence based generative model to generate ECG samples for enhancing emotion state classification. Firstly, emotional related features are extracted to represent emotion state in ECG record. Secondly, random forest and support vector machine are trained to classify arousal and valence states. Then proposed generative model is applied to generate ECG sample with the corresponding emotion state label. Finally, synthetic data is used to augment the original training set for another classification. Our proposed model classifying emotion state in both arousal and valence domain. With synthetic augmented dataset, the average classification accuracy increases around 5% compared with using only original data. The result demonstrates the notable effectiveness of our generative model for enhancing emotion state classification.