| Popis: |
This article presents a method based on artificial neural networks and the Monte Carlo method in which, by using captured signals from conventional mobile phone accelerometers, it is capable of identifying user displacement patterns defined by its localization (inside the pocket, in the hand or on the console) and by the way the user is moving (by car or by foot). Such method enables the identification of risk situations which can culminate in accidents, such as driving or walking while using a smartphone. 3 scenarios are evaluated. Scenario 1 identifies one of 10 possible classes of transport with a 91.6% average hit rate. The second scenario classifies the way the user is moving and results show a 91.6% average hit rate. Scenario 3 uses a hierarchical classifier to identify one of 10 possible classes, achieving the average hit rate of 93.9%. |
