Feature derivation and classification of auditory-motor coupling dynamics in healthy and neurologically impaired adults.

The use of auditory stimuli in rehabilitation to target walking has been evidenced in persons with neurological conditions. The methodologies focus on the synchronisation of persons' steps to auditory stimuli showing that the type of stimuli and tempi significantly affect the synchronisation. However, the dynamic of the interaction over time between the motor system and the auditory stimuli, i.e., when steps are aligned (termed as locking) and not aligned (termed as unlocking) to the beat of the stimuli, remains unclear. Quantifying these dynamics would assist in the development of personalised rehabilitation. Nevertheless, it is currently challenging given the variability of responses per individual over time. We propose a methodological solution to quantify the dynamics of the step-to-beat coupling over time within an experimental paradigm where healthy (n = 7) and neurological impaired (n = 6) participants walk three minutes to music and metronomes at various tempi. We applied window partitioning within the time series to account for the changing pattern. To classify data into locked and unlocked events, features of fluctuation and trend were derived on which two statistical tests (circular statistical test and slope test) were done, respectively. Based on the ground truth, the performance of our proposed method yielded high accuracy (91%), precision (90%) and recall (97%). The standard deviation of the inter-step intervals was then modelled across the label and experimental factors. The proposed method is suitable for quantifying fine-grained observation of the dynamics of auditory-motor coupling in adult healthy and neurological impaired participants, with the potential of designing personalised rehabilitation.