Road-quality Classification and Bump Detection with Bicycle-Mounted Smartphones

The paper proposes a embedded surface road classifier for smartphones used to track and classify routes on bikes. The main idea is to provide, along with the route tracking, information about surface quality of the cycling route (is the surface smooth, rough or bumpy?). The main problem is the quantity of accelerometer data that would have to be uploaded along with GPS track, if the analysis was done off-line. Instead, we propose to classify road surfaces online with an embedded classifier, that has been trained off-line. More specifically, we rely on the accelerometer of a bicycle-mounted smartphone for online classification. We carry out experiments to collect cycling tracks consisting of GPS and accelerometer data, label the data and learn a model for classification, which again is deployed on the smartphone. We report on our experiences with classification accuracy on and runtime performance of the classifier on the smartphone.