Analysing Movement Patterns from Smartphones for the Training of Exoskeletons.

In this paper, an approach to motion analysis that combines two different machine learning techniques, using a single smartphone inside a trouser pocket to record data, is proposed. The resulting program shall be used for data analysis during the use of an exoskeleton and for tracking movements of the lower extremities. The project focuses on the question of whether acceleration data from a smartphone are sufficient to make accurate predictions of movements. A Convolutional Neural Network is used for motion recognition and an agglomerative clustering algorithm is used for a further, more precise analysis of individual movement types. An overall concept and preliminary results of this project are discussed in this work.