The Candrive study aims to improve the current practices of screening elderly drivers in Canada by identifying predictors of motor vehicle collisions from monitoring their daily driving behaviours using in-vehicle sensors. The thesis objective was to characterize the baseline behaviour of stable-health older drivers by proposing parameters of interest for detecting changes in behaviour and methods to differentiate drivers using their maneuvers. The in-vehicle sensor data from 12 stable-health drivers were processed, and a turn-identification algorithm with 97.7% accuracy was created for extracting four maneuvers: accelerating from stop, decelerating to stop, right turns, and left turns on 40 to 60 km/h roadways. Most of the drivers exhibited relatively steady month-to-month acceleration behaviours and lower accelerations in adverse driving conditions, which represented their typical driving behaviours. Drivers can be differentiated by the driving patterns from their maneuvers using a multi-expert classifier, which may be applicable for detecting changes in driving behaviour.