A Domain Generalization and Residual Network-Based Emotion Recognition from Physiological Signals

• DOI: 10.34133/cbsystems.0074
• Publication Date: 2023-11-03T08:12:30Z
• Authors: Junnan Li; Jiang Li; Xiaoping Wang; Xin Zhan; Zhigang Zeng
• Keywords: 03 medical and health sciences; 0302 clinical medicine; Q300-390; Cybernetics

Emotion recognition from physiological signals (ERPS) has drawn tremendous attention and can be potentially applied to numerous fields. Since are nonstationary time series with high sampling frequency, it is challenging directly extract features them. Additionally, there 2 major challenges in ERPS: (a) how adequately capture the correlations between at different times types of (b) effectively minimize negative effect caused by temporal covariate shift (TCS). To tackle these problems, we propose a domain generalization residual network-based approach for emotion (DGR-ERPS). We first pre-extract time- frequency-domain original compose new series. Then, order fully correlation information signals, converted into 3D image data serve as input residual-based feature encoder (RBFE). In addition, introduce generalization-based technique mitigate issue posed TCS. have conducted extensive experiments on real-world datasets, results indicate that our DGR-ERPS achieves superior performance under both TCS non-TCS scenarios.