Abstract
Individualized health-care is gaining traction recently due to the advances in sensor technology, edge computing and the improvement in communication technologies. In this work, we show the application of wearable device for mobile health-care along with supporting algorithms using photoplethysmogram (PPG) as our signal source. We introduce the design of our wearable device equipped PPG sensors for extracting valuable bio-information from the human body. Using our device, we introduce Centralized State Sensing (CSS), an algorithm that improves the estimation of different health parameters. By using heart rate estimation as an example, the proposed algorithm is 26.8% better in terms of mean absolute error when compared with the average reading taken across sensors. We also propose a deep neural network model applied to a PPG-glucose dataset containing 349 samples provided by our industry partner for non-invasive blood glucose estimation achieving an accuracy of 84.29% based on ISO:15197:2013 standard.
| Original language | English |
|---|---|
| Article number | 9389545 |
| Pages (from-to) | 13564-13573 |
| Number of pages | 10 |
| Journal | IEEE Sensors Journal |
| Volume | 21 |
| Issue number | 12 |
| DOIs | |
| State | Published - 15 Jun 2021 |
Keywords
- Neural networks
- non-invasive glucose estimation
- sensor array
