Statistical‐based approach for driving style recognition using Bayesian probability with kernel density estimation

• DOI: 10.1049/iet-its.2017.0379
• Publication Date: 2018-03-08T02:20:51Z
• Authors: Wei Han; Wenshuo Wang; Xiaohan Li; Junqiang Xi
• Keywords: Kernel density estimation; Discriminative model; Kernel (algebra); Feature (linguistics); Feature vector

Driving styles have a great influence on vehicle fuel economy, active safety, and drivability. To recognize driving of path-tracking behaviors for different divers, statistical pattern-recognition method is developed to deal with the uncertainty or characteristics based probability density estimation. First, describe driver styles, speed throttle opening are selected as discriminative parameters, conditional kernel function built, respectively, two representative e.g., aggressive normal. Meanwhile, posterior each element in feature vector obtained using full Bayesian theory. Second, Euclidean distance involved decide which class should be subject instead calculating complex covariance between every elements vectors. By comparing vector, classified into seven levels ranging from low normal high aggressive. Subsequently, show benefits proposed method, cross-validated used, compared fuzzy logic-based method. The experiment results that estimation more efficient stable than