Previous work has shown that a statistical property of gait characterised by long-range autocorrelation functions is altered in Parkinson disease (PD). On the other hand it has been suggested that the same property is linked to the ability in healthy humans to co-regulate the amplitude and cadence of strides towards maintaining a constant speed. Here the investigators want to better understand why it is altered in PD by measuring the transitions between gait instructed by a metronome, and gait without metronome. The experimental conditions will allow the comparisons between these transitions across PD and healthy groups of volunteers, and assess differences based on statistical and computational modelling. The link with potential freezing episodes will also be studied to assess whether the statistical determinants of gait control in this population can be used as a proxy or predictor of the occurence of freezing episodes.