Embedded Emotion Recognition System Based on Electrocardiogram Attributes

Investigated research proved the relevance of the analysis of physiological signals to detect human emotional states. This paper presents an embedded emotion recognition system based on electrocardiogram attributes. We applied the Support Vector Machine (SVM) with a subject independent classification and we implemented the whole proposed system on the Raspberry Pi 3 model B. This System can be easily mounted on robots for an affective Human-Machine interactivity. Thus, we explored the multimodal MAHNOB-HCI database for the two-class problem discrimination in the Arousal-Valence space. After using the ten cross-validations and testing several SVM' kernels, the average classification rates were 57.01% and 54.07% for arousal and valence, respectively.