This work presents the design and evaluation of an activity recognition system for seven important motion related activities. The only sensor used is an Inertial Measurement Unit (IMU) worn on the belt. For classification, we applied Bayesian techniques, based on relevant features of the IMU raw data which are calculated in real time. Based on a complete labelled data set, i.e. supervised by an observing human judge, a K2 learning algorithm by Cooper and Herskovits was used to construct the Bayesian Network (BN) of the features. Our comparison of dynamic and static inference algorithms, based on the evaluation of the labelled data sets recorded from 16 male and female subjects show that a Hidden Markov Model (HMM) based on a learnt BN provides the best results.

This paper compares the performance of inference in static and dynamic Bayesian Networks. For the comparison both kinds of Bayesian networks are created for the exemplary application activity recognition. Probability and structure of the Bayesian Networks have been learnt automatically from a recorded data set consisting of acceleration data observed from an inertial measurement unit. Whereas dynamic networks incorporate temporal dependencies which affect the quality of the activity recognition, inference is less complex for dynamic networks. As performance indicators recall, precision and processing time of the activity recognition are studied in detail.The results show that dynamic Bayesian Networks provide considerably higher quality in the recognition but entail longer processing times.