Authors:
Linda Büker
;
Sandra Hellmers
and
Andreas Hein
Affiliation:
Assistance Systems and Medical Device Technology, Department for Health Services Research, School VI - School of Medicine and Health Sciences, Carl von Ossietzky Universität Oldenburg, Ammerländer Heerstraße 114-118, 26129 Oldenburg, Germany
Keyword(s):
Tremor Classification, Inertial Measurement Units, Support Vector Machine, Activities of Daily Living.
Abstract:
Motor impairments, such as tremors, are often measured with specific tests or rating scales. As these have some disadvantages, like an inter-rater reliability and a lack of representation of the everyday life, a sensor-based continuous and objective monitoring of activities of daily living could be a suitable alternative. According to the literature, the use of inertial measurement units attached to the tremor-dominant arm in combination with support vector machines or neural networks seem to be promising. However, many approaches have to be adapted individually. Therefore, we conducted a preliminary study with ten healthy participants, who were asked to perform conventional and simulated tremor movements during five different activities related to eating. These movements were recorded with inertial measurement units. We identified four different parameters calculated from the recorded data, that we used to train multiple support vector machines for a non-individualized approach. The
overall median accuracy score was 0.75, which is comparable to the results reported in the literature. This shows that support vector machines may be a non-individualized approach for differentiating between tremor and non-tremor movements during activities of daily living.
(More)