Public Transport Driver Identification System Using Histogram of Acceleration Data

Table 1

Comparison between our proposed system and other existing work.

Author	Sensor used for driver identification	Data collection period	Number of drivers	Features	Inactive-driving filtering	Classifier	Accuracy of driver identification (%)	Practicability for impostor detection
Our proposed system with the new data preprocessing	Acceleration sensor	10 months	3-13	Histogram of acceleration	✔	Neural network	94-99	✔; F1 score 0.87 using KNN classifier
Our previous work [19]	Acceleration sensor	10 months	3-13	Histogram of acceleration	✘	Neural network	74-88	✘
Phumphuang et al. [18]	Acceleration sensor	3 months	5	Statistical features with PCA	✘	✘	60-100	✘
Chowdhury et al. [20]	GPS	2 months (1223 hours)	4-5	Statistical features	✘	Random Forest	~ 82.3	✘
Enev et al. [10]	Brake paddle position	~ 3 hours/driver	15	Statistical, descriptive, and frequency features	✘	Random Forest	87-100	✘