Hyper-Parameter Optimization for Emotion Detection using Physiological Signals

摘要

Emotion recognition systems are an automatic way to detect user’s temporary emotional states. Nowadays, there is a high volume of work done regarding the emotion recognition task. However, this task remains to be complex and challenging. In this work, we employed optimization algorithm known as sequential model-based optimization (SMBO) algorithms for segmentation and feature selection to predict affect from physiological signals, such as electrocardiogram and electro-dermal activity, directly from the raw time representation. We presented empirical results for the configuration of a physiological signal-based emotion recognition system. This information may be helpful for people developing emotion recognition systems and for further research in this field.