Dynamic Basic Activity Sequence Matching Method in Abnormal Driving Pattern Detection Using Smartphone Sensors

• DOI: 10.3390/electronics9020217
• Publication Date: 2020-01-27T12:41:11Z
• Authors: Thi-Hau Nguyen; Dang-Nhac Lu; Duc-Nhan Nguyen; Ha-Nam Nguyen
• Keywords: Dynamic Time Warping; Sequence (biology); Weaving

In this work, we present a novel method, namely dynamic basic activity sequence matching (DAS), combination of machine learning methods and flexible threshold based for distinguishing normal abnormal driving patterns. Indeed, DAS relies on the detection module (ADM) presented in our previous work to analyze each pattern as activities—stopping (S), going straight (G), turning left (L), right (R). fact, value other parameters like duration long short activities are iteratively induced from collected dataset. Hence, is independent contexts such vehicle modes road conditions. Experimental results, dataset numerous motorcyclists, show outperformance proposed method against time warping two popular methods—random forest neural network—in Moreover, propose an efficient framework composing phases: first phase, patterns distinguished by relying DAS. second detected further classified into various specific patterns—weaving, sudden braking, etc. This fusion again achieves highest overall accuracy 97.94%.