Experiments on decision fusion for driver recognition

In this work, we study the individual as well as combined performance of various driving behavior signals on identifying the driver of a motor vehicle. We investigate a number of classifier fusion techniques to combine multiple channel decisions. We observe that some driving signals carry more biometric information than others. When we use trainable combining methods, we can reduce identification error significantly using only driving behavior signals. Classifier combination methods seem to be very useful in multi-modal biometric identification in a car environment.