Emotion Detection Using EEG and ECG Signals from Wearable Textile Devices for Elderly People

The global population is ageing; exacerbating a range of age-related health problems, like dementia. In the late stage of dementia, patients often are unable to find words to express their feelings; causing serious challenges in healthcare. Our aim is to detect the emotions of elderly patients using physiological signals - electroencephalogram (EEG) and electrocardiogram (ECG) - using deep learning neural networks. However, most EEG and ECG monitoring devices are uncomfortable and not suitable for daily wear by elderly people. For this study, a prior experiment was conducted on 5 healthy elderly subjects for binary classification of positive and negative emotions: EEG and ECG data were collected from the subjects, using our own designed wearable textile devices while they watch selected stimuli. We propose an end-to-end deep learning method - Long short-term memory (LSTM) - to detect emotion from raw clean signals after removing noises and baseline wander. LSTM can learn features from raw data directly and achieve binary emotion classification with an accuracy of 76.67% with EEG signals, 75.00% with ECG signals, and 95.00% with EEG and ECG signals, respectively. This proposed system for detecting emotion by deep learning method using our user-friendly and easy-to-wear textile devices offer great prospects for use in everyday care situations and dementia care.