Driver sleepiness detection in real driving situations, a pre-study
| Reference number | |
| Coordinator | CHALMERS TEKNISKA HÖGSKOLA AKTIEBOLAG - SAFER & Signaler & System |
| Funding from Vinnova | SEK 707 000 |
| Project duration | March 2015 - December 2015 |
| Status | Completed |
| Venture | Traffic safety and automated vehicles -FFI |
| End-of-project report | 2014-06235eng.pdf (pdf, 332 kB) |
Purpose and goal
Developing and algorithm that accurately discriminates between awake and sleepy driver was achieved.
Results and expected effects
The outcome achieved with this project together with unobtrusive accurate HR detection are the two necessary pieces of a driver sleepiness detection system based on physiological measurements.
Approach and implementation
The ECG signal w as divided in epochs of 5 minutes, corresponding to the occurrences of the KSS in time. The R peaks detection was done using the Pan-Tompkins algorithm. Outliers detection was based on a heart IBI differing in more than 30% from the mean value of the four previous intervals; spectral transformation was based on the Fourier transform. Most often HRV indices used were included as predictors for the classifier. This classifier was based a SVM with radial kernel. Ten-fold cross-validation was the testing method of choice.