#$%^&*AU2014277815B220170302.pdf#####ABSTRACT An automated method of (and device for) determining a dyskinetic state of a person, the method comprising: obtaining a time series of accelerometer data, from an accelerometer worn on an extremity of the person, over an extended period during everyday activities of the person； band pass filtering the accelerometer data to extract filtered data for a band of interest, wherein the band of interest has a lower end cut off frequency selected to remove DC and an upper end cut off frequency which is selected to eliminate higher frequency components which in general do not arise from normal human motions； extracting one or more bins of acceleration data from the time series of the acceleration data, each bin being of a time duration which is selected to be small enough that relatively regular measures of dyskinesia are determined, while being long enough to provide a reasonable likelihood of a significant movement by the person during that bin； comparing the accelerometer data to a threshold value； determining a proportion of time for which the accelerometer data remains below the threshold value, where the determined proportion of time is representative of a time of reduced movement (TRM)； and processing the filtered data and the TRM to determine a measure for dyskinesia.
