Nonlinear methods to quantify Movement Variability in Human-Humanoid Interaction Activities

Human movement variability arises from the process of mastering redundant (bio)mechanical degrees of freedom to successfully accomplish any given motor task where flexibility and stability of many possible joint combinations helps to adapt to environment conditions. While the analysis of movement of variability is becoming increasingly popular as a diagnostic tool or skill performance evaluation, there are remain challenges on applying the most appropriate methods. We therefore investigate nonlinear methods such as reconstructed state space (RSSs), uniform time-delay embedding, recurrence plots (RPs) and recurrence quantification analysis (RQAs) with real-world time-series data of wearable inertial sensors. That said, twenty healthy participants imitated vertical and horizontal arm movements in normal and faster velocity from an humanoid robot. We applied nonlinear methods to the collected data to found visual differences in the patterns of RSSs and RPs and statistical differences with RQAs. We conclude that Shannon Entropy with RQA is a robust method that helps to quantify activities, types of sensors, windows lengths and level of smoothness. Hence this work might enhance the development of better diagnostic tools for applications in rehabilitation and sport science for skill performance or new forms of human-humanoid interaction for quantification of movement adaptations and motor pathologies.