Developing a motion language: Gait analysis from accelerometer sensor systems

This work concerns the use of human movement classification as a tool for monitoring and supporting older peoples' lives. The “motion language” methodology is a movement classification technique which aims at generalizing movements and providing easy interpretation of motion signals by decomposing activities into elementary building blocks referred to as “motion primitives”. The use of motion primitives to classify motion from visual data has been studied. This work shows that the motion language methodology can be applied to acceleration signals, contributing to the development of wearable monitoring systems. This paper explains the development of the motion language and its use in a gait analysis study. Preliminary results show that the motion language methodology can be used to quantitatively measure gait parameters. In addition, motion primitives are shown to express static and dynamic characteristics of different gait patterns and were used to calculate a new symmetry index.