An R package housing commonly used signal based feature extraction methods for machine learning algorithm development. It is built with IMU signals (accelerometer, gyroscope) in mind, however, many of the features will work for any signal.
The package contains the following features
-
Mean
-
standard deviation (a.k.a root mean square)
-
inter-quartile range
-
skewness
-
kurtosis
-
ratio_beyond_r_sigma: percent of observation that is outside r standard deviations
-
Slope for linear trend
-
autogressive coefficient lag 1
-
autocorrelation lag 1
-
mean cross rate
-
entropy
-
sample entropy
-
permutation entropy
-
dominant frequency
-
dominant frequency magnitude
-
spectral flatness
-
spectral entropy
-
jerk
-
sparc
-
dimensionless jerk
-
log dimensionless jerk