On the structure of natural human movement

Understanding of human motor control is central to neuroscience with strong implications in the fields of medicine, robotics and evolution. It is thus surprising that the vast majority of motor control studies have focussed on human movement in the laboratory while neglecting behaviour in natural environments. We developed an experimental paradigm to quantify human behaviour in high resolution over extended periods of time in ecologically relevant environments. This allows us to discover novel insights and contradictory evidence to well-established findings obtained in controlled laboratory conditions. Using our data, we map the statistics of natural human movement and their variability between people. The variability and complexity of the data recorded in these settings required us to develop new tools to extract meaningful information in an objective, data-driven fashion. Moving from descriptive statistics to structure, we identify stable structures of movement coordination, particularly within the arm-hand area. Combining our data with numerous published findings, we argue that current hypotheses that the brain simplifies motor control problems by dimensionality reduction are too reductionist. We propose an alternative hypothesis derived from sparse coding theory, a concept which has been successfully applied to the sensory system. To investigate this idea, we develop an algorithm for unsupervised identification of sparse structures in natural movement data. Our method outperforms state-of-the-art algorithms for accuracy and data-efficiency. Applying this method to hand data reveals a dictionary of \emph{sparse eigenmotions} (SEMs) which are well preserved across multiple subjects. These are highly efficient and invariant representation of natural movement, and suggest a potential higher-order grammatical structure or ``movement language''. Our findings make a number of testable predictions about neural coding of movement in the cortex. This has direct consequences for advancing research on dextrous prosthetics and robotics, and has profound implications for our understanding of how the brain controls our body.Open Acces