Feedback and inquiries

Intelligent Systems Group
Department of Computer Science and Engineering P.O.Box 4500
FIN-90014 University of Oulu
Finland
Tel. +358-8-553 1011
Fax. +358-8-553 2612

Context recognition

Our aim is to find and develop methods for recognizing the user’s context from sensor data using signal processing, pattern recognition and machine learning methods. Automatic recognition of context enables building easy to use applications that adapt as the user context changes and applications that simply collect and visualize information, providing means for decision making. Sensor data processing is needed for physical user interfaces as well.

In our earlier research, we have developed methods for recognizing context from location and acceleration data produced by wearable sensors. In 2003, we detected the phone profiles the user selects at different places; results for one user are shown below. As another example of our earlier work, in 2006 we studied, in collaboration with Waseda University, recognizing the activities a user is performing from data produced by wearable acceleration sensors.

Currently we focus on traffic and road condition analysis using sensors embedded in mobile phones. Information about not only the road condition but also about the types of road users such as pedestrians and cyclists, for example, can be useful and may prevent accidents. Warning about an approaching pedestrian in dark or a large bump on a main road are some exemplars.

Road condition information is seen useful for both road users as well as for the road network operators. Our purpose is to develop a signal processing and classification system for recognizing road condition from accelerometer and GPS readings. The goal is to prove the feasibility of road condition monitoring using participatory sensing with mobile phones. Our first prototype is shown below. This prototype was built in collaboration with University of Jyväskylä, Tampere University of Technology, and Department of Information Processing Science from our University (in the SDFA project). Mobile phones attached to vehicles collect acceleration and GPS data and send the data to the server. Road condition is recognized from this data. Clients can request data from the server, for example, to visualize it for the road users.

The second image shows how the phone is attached to a vehicle, and gives some details about processing.

We also study recognition of context from ambient audio. Here, audio from users’ everyday environments is recorded using a mobile device and through signal processing and classification used as a cue of the users’ current context. The architecture of the recognition system is shown below. This system was implemented on a mobile phone.

As part of the context recognition work, we are building a system for collecting, processing, storing, and disseminating sensor data. We selected the Global Sensor Network platform as the basis and are currently developing data processing modules for that platform. We have built the first prototypes (in 2009-2010) for traffic and road condition analysis and for collecting data from a wireless sensor network. Our first wireless sensor network prototype is shown below. This prototype was built in the RealUbi project; our work focused on the GSN.