Nonlinear analysis to quantify human movement variability from time-series data

The complexity of human movement arises from the management of redundant (bio)mechanical degrees of freedom (DOF) to successfully accomplish any given motor task (Bernstein, 1967). Such DOF can be described in terms of (in)dependent coordinates which uniquely describe a joint configuration of the motor system. The joint configuration provides flexibility and stability to allow consistent movement, and the management of DOF allow adaptation to changing environment conditions. Consequently, human movement variability is always present in even the simplest movements. That said, this talk will touch on three points: (a) theoretical models for human movement variability and methods to quantify human movement variability, (b) use of nonlinear methods to measure real-world time series data (i.e., data affected by non-stationarity, non-linearity, data length, sensor source, noise, etc.), and (c) illustration of results for real-world time-series data from human-robot imitation activities. I will then comment on current and future challenges on this subject and how the above points might lead to development tools to evaluate, for instance, the improvement of movement performances, to quantify and provide feedback of skill learning or to quantify movement adaptations and pathologies.