A Study on Driving Behavior Intelligence Detection Based on Discrete Wavelet Transform and Support Vector Machine Algorithm

With the rapid development of Artificial Intelligence, big data analysis, smart detection of driving behaviors becomes the new focus of Intelligence and Connected Vehicles researches. The state-of-the-art research in this direction is to recognize driving scenes to support driving decisions, based on driver's driving data. This study used CATARC collected Controller Area Network (CAN) bus and CARTAC driving scene standard labeled data, implemented discrete wavelet transform (DWT) and support vector machine (SVM) algorithm, constructed a machine learning model with the ability of detecting 16 different driving behaviors. With details of feature selection, filter selection, SVC parameters selection and many others techniques to optimize the model, achieved cross validation accuracy rate around 88%. This method can be applied to vehicles' security warnings and intelligence control, therefore to improve vehicle safety performance.