Exploring temporal representations by leveraging attention-based bidirectional LSTM-RNNs for multi-modal emotion recognition
作者：
Highlights：
• Compared with hand-craft features, the proposed method can automatically extract temporal features from raw physiological signals by Attention-based BLSTM-RNNs, which is capable of learning feature representations and modeling the temporal dependencies between their activation.
• We also investigate the usage of attention-based architectures to improve physiological-based emotion recognition. The attention mechanism allows the network to focus on the emotionally salient parts of a sequence.
• Decision level fusion is implemented to capture complementary information from different modalities for enhancing emotion recognition system.
摘要
•Compared with hand-craft features, the proposed method can automatically extract temporal features from raw physiological signals by Attention-based BLSTM-RNNs, which is capable of learning feature representations and modeling the temporal dependencies between their activation.•We also investigate the usage of attention-based architectures to improve physiological-based emotion recognition. The attention mechanism allows the network to focus on the emotionally salient parts of a sequence.•Decision level fusion is implemented to capture complementary information from different modalities for enhancing emotion recognition system.
论文关键词：Emotion recognition,EEG signals,Physiological signals,Deep learning,Multimedia content,Multi-modal fusion,00-01,99-00
论文评审过程：Received 29 July 2019, Revised 15 November 2019, Accepted 16 December 2019, Available online 3 January 2020, Version of Record 3 January 2020.
论文官网地址：https://doi.org/10.1016/j.ipm.2019.102185
