Driver Identification by an Ensemble of CNNs Obtained from Majority-Voting Model Selection

Driver identification refers to the task of identifying the driver behind the wheel among a set of drivers. It is applicable in intelligent insurance, public transportation control systems, and the rental car business. An critical issue of these systems is the level of privacy, which encourages a lot of research using non-visual data. This paper proposes a novel method based on IMU sensors’ data of smartphones. Also, an ensemble of convolutional neural networks (CNNs) is applied to classify drivers. Furthermore, the final prediction is obtained by a majority vote mechanism. This paper demonstrates that model selection using a majority vote significantly improves the accuracy of the model. Finally, the performance of this research in terms of the accuracy, precision, recall, and f1-measure are 93.22%, 95.61%, 93.22%, and 92.80% respectively when the input length is 5 min.