Scholarlink

Emotion recognition from physiological signals is pivotal for enhanced human–computer interaction. However, recognizing emotions using conventional electroencephalogram (EEG) signals offers limited wearability in day-to-day scenarios. In addition, the recent advancements in wearable fitness tracking devices offer non-intrusive, real-time insights into individuals’ health parameters. Thus, modeling diverse multimodal physiological modalities to recognize human emotions can enable an emotionally enriched user experience. The work proposes an automated emotion recognition system utilizing the multimodal physiological signals acquired using the wearable smartwatch. The study utilizes the publicly available EMOGNITION database comprising diverse physiological parameters. The parameters considered in this work include Heart Rate, Blood Volume Pulse, 3-axis Accelerometer, 3-axis Gyroscope, and 4-axis Rotation, acquired via the Samsung Galaxy smartwatch wearable device. The work examines the variations in the considered signals to determine nine discrete emotions. The preprocessed signal modalities are investigated using the hybrid Convolutional Neural Networks (CNNs) and one layer of Long Short-term Memory (LSTM) deep learning (DL) architecture. The adopted DL architecture obviated the requirement of manual feature extraction. The developed system successfully determines nine discrete emotions with an average accuracy and F1 score of 96.82% and 96.74%, respectively. The obtained results outperform the existing methods in the literature, demonstrating potential in emotion recognition.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrated CNN-LSTM Model for Emotion Detection Using Physiological Signals from Wearables

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ankita,
  • Inderveer Singh,
  • Durgesh Nandini,
  • Jyoti Yadav

摘要

Emotion recognition from physiological signals is pivotal for enhanced human–computer interaction. However, recognizing emotions using conventional electroencephalogram (EEG) signals offers limited wearability in day-to-day scenarios. In addition, the recent advancements in wearable fitness tracking devices offer non-intrusive, real-time insights into individuals’ health parameters. Thus, modeling diverse multimodal physiological modalities to recognize human emotions can enable an emotionally enriched user experience. The work proposes an automated emotion recognition system utilizing the multimodal physiological signals acquired using the wearable smartwatch. The study utilizes the publicly available EMOGNITION database comprising diverse physiological parameters. The parameters considered in this work include Heart Rate, Blood Volume Pulse, 3-axis Accelerometer, 3-axis Gyroscope, and 4-axis Rotation, acquired via the Samsung Galaxy smartwatch wearable device. The work examines the variations in the considered signals to determine nine discrete emotions. The preprocessed signal modalities are investigated using the hybrid Convolutional Neural Networks (CNNs) and one layer of Long Short-term Memory (LSTM) deep learning (DL) architecture. The adopted DL architecture obviated the requirement of manual feature extraction. The developed system successfully determines nine discrete emotions with an average accuracy and F1 score of 96.82% and 96.74%, respectively. The obtained results outperform the existing methods in the literature, demonstrating potential in emotion recognition.