Physiological sensor analytics aims at monitoring health as the availability of sensor-enabled portable, wearable, and implantable devices become ubiquitous in the growing Internet of Things (IoT). Physiological multi-sensor studies have been conducted previously to detect stress. In this study, we focus on electrocardiography (ECG) monitoring that can now be performed with minimally invasive wearable patches and sensors, to develop an efficient and robust mechanism for accurate stress identification, for example in automobile drivers. A unique aspect of our research is personalized individual stress analysis including three stress levels: low, medium and high. Using machine learning algorithms from the ECG signals alone, our system achieves up to 100% accuracy and area under ROC curve of 1 depending on the experimental setting in detecting three classes of stress using feature selection from a combination of fiducial points and multiscale entropy as a fine-grained indicator of stress level.
CITATION STYLE
Bichindaritz, I., Breen, C., Cole, E., Keshan, N., & Parimi, P. (2018). Feature selection and machine learning based multilevel stress detection from ECG signals. In Smart Innovation, Systems and Technologies (Vol. 71, pp. 202–213). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-319-59397-5_22
Mendeley helps you to discover research relevant for your work.