BROWSE
Detail
Conference Paper
Attention-LRCN: Long-term Recurrent Convolutional Network for Stress Detection from Photoplethysmography
Cited 9 time in 
Share 
Share
- Authors
- Jiho Choi, Jun Seong Lee, Moonwook Ryu, Sang Jun Lee
- Issue Date
- 2022-06
- Citation
- International Symposium on Medical Measurements and Applications (MeMeA) 2022, pp.1-5
- Language
- English
- Type
- Conference Paper
- Abstract
- Recently, interest in well-being has been increasing rapidly, and one way to do this is to deal with stress wisely. In order to manage or relieve stress, it is necessary to identify the current stress status and respond appropriately. Many existing studies have been conducted to detect stress, and lately many deep learning-based stress detection methods have been proposed. However, there is a room for improving the accuracy, and this paper proposes a novel deep learning algorithm for stress detection. The proposed model is based on long-term recurrent convolutional networks (LRCN) and an attention module, and we named this as Attention-LRCN. We used WESAD dataset which provides photoplethysmography (PPG) signals with normal and stress statuses for 15 subjects. The proposed method classifies the PPG signal into stress and normal statuses using a combination of convolutional neural networks (CNN) and long short-term memory (LSTM) layers. Since the PPG signals contain human interference, we utilized an attention module to reduce the effects of noise on the PPG signal. We compare Attention-LRCN with the state-of-the-art method for stress detection, and experimental results demonstrate that our proposed method is more effective in the stress detection application. The proposed method achieved 97.11 % and 95.47% for the accuracy and F1-score, respectively, and these metrics are 0.61 % and 2.1 % higher than the state-of-the-art method.
- KSP Keywords
- Convolution neural network(CNN), Convolutional networks, Detection Method, F1-score, Learning-based, Long-short term memory(LSTM), PPG signal, Proposed model, Stress detection, current stress, deep learning(DL)