In-Vehicle Activity Recognition of User Activities Using Smartwatches

Hald, Thomas Alexander Cano ; Mårtensson, Mads

4. term

Computer Science, Master

2017

2017-06-08

42 pages

In this thesis, we investigate the use of motion sensors in a commercially available smartwatch as a sensing unit in an in-vehicle role identification system (IRIS). Our contribution is two-fold. Firstly, a system to provide mobile applications with in-vehicle contextual information about a user's role in a car, e.g. driver or passenger, and evaluation of what influences it, e.g. routes and roads. Secondly, an application, Hands-On, that leverages such contextual information to determine whether a driver has their hands in a recommended hand position and exploration of a small group of users' initial reactions to such an application. With regards to mobile devices, an ongoing challenge is designing for context as mobile contexts are highly dynamic and complex. Previously, mobile applications have been made orientation aware and aware of basic daily activities, such as walking and running. The use of the motion sensors in these devices have shown potential for making the mobile device context aware on a deeper activity level through research in the area known as Human Activity Recognition. Within this area, the recognition in-vehicle activities has been investigated.

Smartwatches ; Human Activity Recognition ; Machine Learning ; In-vehicle systems

Download
View record in AAU Student Projects