Abstract
People increasingly interact with social media or other apps on their smartphones while driving car. This is naturally a major safety concern, and it remains unclear how to avoid or limit such interaction. We investigate this problem through human activity recognition (HAR) where we developed a system called IRIS, which collects smartwatch accelerometer data and analyses the data through machine learning and predicts if the data origins from a driver or a passenger. We report from a field experiment with 24 participants acting as drivers or passengers where we achieved an overall prediction accuracy of 87%. We further found that various road segments had less effect on the accuracy than anticipated, but we also found that passenger tasks had a negative effect on recognition accuracy. We discuss several implications from findings.
Original language | English |
---|---|
Title of host publication | Proceedings of the 10th International Conference on Automotive User Interfaces and Interactive Vehicular Applications |
Number of pages | 11 |
Publisher | Association for Computing Machinery |
Publication date | 23 Sept 2018 |
Pages | 74-84 |
ISBN (Electronic) | 978-1-4503-5946-7 |
DOIs | |
Publication status | Published - 23 Sept 2018 |
Event | 10th International ACM Conference on Automotive User Interfaces - Toronto, Canada Duration: 23 Sept 2018 → 25 Sept 2018 |
Conference
Conference | 10th International ACM Conference on Automotive User Interfaces |
---|---|
Country/Territory | Canada |
City | Toronto |
Period | 23/09/2018 → 25/09/2018 |
Keywords
- Accelerometer data
- Driving
- Human activity recognition
- Sensor data
- Smartwatch