Table 4.

Feature characteristics. Feature characteristics computed for each individual window. Here f represents the signal and fs the sampling frequency.

feature characteristics	description/formula
interquartile range	difference between the 75th percentile and the 25th percentile value of a window
kurtosis	kurtosis calculated from window values
mean	mean of all window values
standard deviation	standard deviation of all window values
minimum value	minimum value of all window values
maximum value	maximum value of all window value
number of zero crossings	number of zero crossings in a window after subtracting the window mean value from every window sample
spectral entropy	power spectral density:
	PSD = |X(f)2|
	X(f) – DFT of original signal. Discrete Fourier transform (DFT)
	normalized PSD:
	${\text{PSD}}_{\mathrm{norm}}(f)={\displaystyle \frac{\mathrm{PSD}(f)}{{\mathrm{\Sigma }}^{\mathrm{PSD}}(f)}}$
	spectral entropy:
	$\text{SE}=-{\mathrm{\Sigma }}_{\mathrm{norm}}^{\mathrm{PSD}}(f)\cdot \log (\text{PS}{\text{D}}_{\mathrm{norm}}(f))$
dominant frequency	after applying Fourier transformation, this is the frequency at which the signal has its highest power
signal area	signal area:
	$\text{SA}=\mathrm{\Sigma }\text{Mag}\cdot \frac{1}{fs}$
	Mag—acceleration or gyroscope magnitude
	fs—sampling frequency
absolute signal area	absolute signal area:
	$\text{ASA}=\mathrm{\Sigma }|\text{Mag}|\cdot \frac{1}{fs}$