Creating and Benchmarking a New Dataset for Physical Activity Monitoring

Creating and Benchmarking a New Dataset for Physical Activity Monitoring
Attila Reiss, Didier Stricker
Proceedings of the PETRA 2012 Conference International Conference on Pervasive Technologies Related to Assistive Environments (PETRA-2012), The 5th Workshop on Affect and Behaviour Related Assistance (ABRA), June 6-8, Crete, Greece

Abstract:
Physical activity monitoring has recently become an important field in wearable computing research. However, there is a lack of a commonly used, standard dataset and established benchmarking problems. In this work, a new dataset for physical activity monitoring - recorded from 9 subjects, wearing 3 inertial measurement units and a heart rate monitor, and performing 18 different activities - is created and made publicly available. Moreover, 4 classification problems are benchmarked on the dataset, using a standard data processing chain and 5 different classifiers. The benchmark shows the difficulty of the classification tasks and exposes some challenges, defined by e.g. a high number of activities and personalization.

Creating and Benchmarking a New Dataset for Physical Activity Monitoring

Creating and Benchmarking a New Dataset for Physical Activity Monitoring
Attila Reiss, Didier Stricker
Proceedings of the PETRA 2012 Conference International Conference on Pervasive Technologies Related to Assistive Environments (PETRA-2012), The 5th Workshop on Affect and Behaviour Related Assistance (ABRA), June 6-8, Crete, Greece

Abstract:
Physical activity monitoring has recently become an important field in wearable computing research. However, there is a lack of a commonly used, standard dataset and established benchmarking problems. In this work, a new dataset for physical activity monitoring - recorded from 9 subjects, wearing 3 inertial measurement units and a heart rate monitor, and performing 18 different activities - is created and made publicly available. Moreover, 4 classification problems are benchmarked on the dataset, using a standard data processing chain and 5 different classifiers. The benchmark shows the difficulty of the classification tasks and exposes some challenges, defined by e.g. a high number of activities and personalization.