Utilization of smartphone sensor data for driving state classification

Abstract

Numerous experiments were carried out using a car driving into a multi-storey carpark attached to a shopping mall. The dataset was collected using accelerometer sensor embedded in a smartphone which was placed in the car during the experiment. The collected data can be categorised into driving, idling and walking. The main focus of this project is to identify different motion states occurred in the parking session. Two popular classifiers K-Nearest Neighbour and Support Vector Machine have been evaluated using various parameters to achieve optimal performance. Features were also extracted from the raw dataset to improve classification accuracy.