Skip to main content
. 2017 May 25;19(5):e184. doi: 10.2196/jmir.7385

Table 1.

Activity recognition model features from accelerometer and gyroscope signals.

Description Number of features
Mean, range, and interquartile range 3 (per axis)
Moments: standard deviation, skew, and kurtosis 3 (per axis)
Histogram: bin counts of z-scores from -2 to 2 4 (per axis)
Moments of derivative: mean, standard deviation (SD), skew, and kurtosis 4 (per axis)
Mean of the squared norm 1
Sum of axial standard deviations 1
Pearson correlation coefficient, r(xy), r(xz), r(yz) 1 (per axis)
Mean cross products (raw and normalized), xy, xz, yz 2 (per axis)
Absolute mean of cross products (raw and normalized) 2 (per axis)
Power spectra: mean, standard deviation, skew, and kurtosis 4 (per axis)
Mean power in 0.5 Hz bins from 0-10 Hz 20 (per axis)