Smartphone-Based Systems for Driving Evaluation

By: Mehdi Ghatee

Published in: Smartphones: Recent Innovations and Applications, Nova Science Pub Inc, United States, (August 15, 2019).


Smartphones play important roles in intelligent transportation systems and driving behavior evaluation contexts. By data mining of smartphone data, the driving style can be recognized efficiently. Then any incentive or punitive mechanisms can be applied to encourage the drivers to perform better. The insurance and enforcement departments can easily use smartphone-based systems to check the driving behaviors gradually. Since, a smartphone consists of many sensors, memory and a high processing power, it can be used as a black-box for storing the driver statue. When an agency needs to understand something about driving or crashing, the smartphone data can be analyzed to clear the situations. However, such data should be collected with low energy consumption. Tothis aim, the multiple criteria such as precision, frequency, energy consumption, data quality, security and privacy are factors to select a reasonable set of sensors for data collection and data mining. In this problem, the type of communication is also important. For preprocessing of these data, different filters are developed to clean data and remove noise and outliers. For sensors fusion and to model data, some neural networks or other machine learning algorithms can be applied. Then, the data are compressed by some intervals, Fourier transforms, wavelets transforms, statistical distribution, stochastic variables or fuzzy quantities. To extract the knowledge from these data, the necessary features can be extracted from the raw data by the aid of matrix decomposition, PCA, deep learning, etc. Sometimes, it is necessary to reduce the dimension of features by feature selection methods. For processing, based on the needs, we can implement data visualizers, predictors, classifiers, frequent pattern miners or clusters extractors. We discuss on the different concepts of data mining approaches on smartphone data in this chapter. Some applications based on these processed data are developed to warn drivers or pedestrians, to define insurance encouragement, or to provide a background for autonomous driving.


smartphone sensors, data mining, sensor fusion, feature extraction, feature selection, machine learning, driving profiling

Leave a comment

Your email address will not be published. Required fields are marked *