Role of data measurement characteristics in the accurate detection of Parkinson’s disease symptoms using wearable sensors

Table 3 Feature categorization for supervised machine learning models

Feature category	Abbreviation	Features	No. Tri-axial features	No. Magnitude features
Time	T	Root mean square, range, mean, variance, skew, kurtosis	18	6
Frequency	F	Dominant frequency, Relative magnitude, Moments of power spectral density (mean, standard deviation, skew, kurtosis)	18	6
Entropy	E	Sample entropy	3	1
Correlation	C	Cross-correlation peak (XY,XZ,YZ), Cross-correlation lag (XY,XZ,YZ)	6	0
Derivative	D	Moments of the signal derivative (mean, standard deviation, skew, kurtosis)	12	4
*Total for each sensor type*			57	17

Features extracted from both accelerometer and gyroscope data signals and used as inputs for symptom models. Features are shown split into the categories used during the analysis of feature types

ISSN: 1743-0003