Recognizing complex upper extremity activities using body worn sensors

Abstract

To evaluate arm-hand therapies for neurological patients it is important to be able to assess actual arm-hand performance objectively. Because instruments that measure the actual quality and quantity of specific activities in daily life are lacking, a new measure needs to be developed. The aims of this study are to a) elucidate the techniques used to identify upper extremity activities, b) provide a proof-of-principle of this method using a set of activities tested in a healthy adult and in a stroke patient, and c) provide an example of the method’s applicability in daily life based on readings taken from a healthy adult. Multiple devices, each of which contains a tri-axial accelerometer, a tri-axial gyroscope and a tri-axial magnetometer were attached to the dominant hand, wrist, upper arm and chest of 30 healthy participants and one stroke patient, who all performed the tasks ‘drinking’, ‘eating’ and ‘brushing hair’ in a standardized environment. To establish proof-of-principle, a prolonged daily life recording of 1 participant was used to identify the task ‘drinking’. The activities were identified using multi-array signal feature extraction and pattern recognition algorithms and 2D-convolution. The activities ‘drinking’, ‘eating’ and ‘brushing hair’ were unambiguously recognized in a sequence of recordings of multiple standardized daily activities in a healthy participant and in a stroke patient. It was also possible to identify a specific activity in a daily life recording. The long term aim is to use this method to a) identify arm-hand activities that someone performs during daily life, b) determine the quantity of activity execution, i.e. amount of use, and c) determine the quality of arm-hand skill performance.