Recognition of Elementary Upper Limb Movements in an Activity of Daily Living Using Data from Wrist Mounted Accelerometers

In this paper we present a methodology as a proof-of-concept for recognizing fundamental movements of the human arm (extension, flexion and rotation of the forearm) involved in 'making-a-cup-of-tea', typical of an activity of daily-living (ADL). The movements are initially performed in a controlled environment as part of a training phase and the data are grouped into three clusters using k-means clustering. Movements performed during ADL, forming part of the testing phase, are associated with each cluster label using a minimum distance classifier in a multi-dimensional feature space, comprising of features selected from a ranked set of 30 features, using Euclidean and Mahalonobis distance as the metric. Experiments were performed with four healthy subjects and our results show that the proposed methodology can detect the three movements with an overall average accuracy of 88% across all subjects and arm movement types using Euclidean distance classifier.