Measuring devices as well as statistical machine learning methods have rapidly developed during the recent years. This has enabled the utilization of small-scale wearable and embedded sensors in context recognition of human motion and his activities.

Jaakko Suutala presents in his PhD thesis novel statistical machine learning and pattern recognition methods, which are applied to sensor-based human context recognition. More precisely, Suutala´s thesis concentrates on an effective discriminative learning framework, where input-output mapping is learned directly from a labeled dataset.

Suutala developed two novel methods for handling sequential input and output data: a novel kernel based on graphical presentation called a weighted walk-based graph kernel (WWGK), and discriminative temporal smoothing (DTS). The proposed algorithms are modular, so different kernel classifiers can be used as base models. Finally, the thesis presents a group of techniques based on Gaussian process regression (GPR) and particle filtering (PF) to learn to track multiple targets.

The proposed methodology is applied to three different human-motion-based context recognition applications: person identification, person tracking, and activity recognition, where floor (pressure-sensitive and binary switch) and wearable acceleration sensors are used to measure human motion and gait during walking and other activities.

The new methods for context recognition can be used in developing new services and applications for intelligent spaces, robotics, mobile computing, or health and well-being.

Jaakko Suutala will defend his PhD thesis on Wednesday, June 27, at 12 noon in lecture room TS 101, Tietotalo I.