The following features used were extracted using mhealthtools [https://github.com/Sage-Bionetworks/mhealthtools].
| Feature Name | Definition |
|---|---|
| meanTapInter | mean tapping interval |
| medianTapInter | median tapping interval |
| iqrTapInter | interquartile range tapping interval |
| minTapInter | minimum tapping interval |
| maxTapInter | maximum tapping interval |
| skewTapInter | skewness tapping interval |
| kurTapInter | kurtosis tapping interval |
| sdTapInter | standard deviation tapping interval |
| madTapInter | mean absolute deviation tapping interval |
| cvTapInter | coefficient of variation tapping interval |
| rangeTapInter | range tapping interval |
| tkeoTapInter | teager-kaiser energy operator tapping interval |
| ar1TapInter | autocorrelation (lag = 1) tapping interval |
| ar2TapInter | autocorrelation (lag = 2) tapping interval |
| fatigue10TapInter | difference in mean tapping interval between the first and last 10% of the tapping interval series |
| fatigue25TapInter | difference in mean tapping interval between the first and last 25% of the tapping interval series |
| fatigue50TapInter | difference in mean tapping interval between the first and last 50% of the tapping interval series |
| meanDriftLeft | mean drift in the left button |
| medianDriftLeft | median drift in the left button |
| iqrDriftLeft | interquartile range of drift in the left button |
| minDriftLeft | minimum of drift in the left button |
| maxDriftLeft | maximum of drift in the left button |
| skewDriftLeft | skewness of drift in the left button |
| kurDriftLeft | kurtosis of drift in the left button |
| sdDriftLeft | standard deviation of drift in the left button |
| madDriftLeft | mean absolute deviation of drift in the left button |
| cvDriftLeft | coefficient of variation of drift in the left button |
| rangeDriftLeft | range of drift in the left button |
| meanDriftRight | mean drift in the right button |
| medianDriftRight | median drift in the right button |
| iqrDriftRight | interquartile range of drift in the right button |
| minDriftRight | minimum of drift in the right button |
| maxDriftRight | maximum of drift in the right button |
| skewDriftRight | skewness of drift in the right button |
| kurDriftRight | kurtosis of drift in the right button |
| sdDriftRight | standard deviation of drift in the right button |
| madDriftRight | mean absolute deviation of drift in the right button |
| cvDriftRight | coefficient of variation of drift in the right button |
| rangeDriftRight | range of drift in the right button |
| numberTaps | number of taps |
| buttonNoneFreq | frequency where neither the left or right buttons were hit |
| corXY | correlation between the X and Y coordinates of the hits |
The following features used were extracted using mpowertools [https://github.com/itismeghasyam/mpowertools].
The following features are of the format <feature><axis> - for example meanX, mean of X axis ; iqrAA, inter quartile range of AA axis (average acceleration)
| <feature> | Definition |
|---|---|
| mean | mean |
| sd | standard deviation |
| mode | mode |
| skew | skewness |
| kur | kurtosis |
| q1 | first quartile cutoff |
| median | median |
| q3 | third quartile cutoff |
| iqr | interquartile range |
| range | range(maximum - minimum) |
| acf | autocorrelation ( at lag = 1) |
| zcr | zero-crossing rate |
| dfa | scaling exponent of the detrended fluctuation analysis of the given series |
| cv | coefficient of variation of the given series |
| tkeo | teager-kaiser energy operator of the given series |
| F0 | frequency at which the maximum peak of the Lomb-Scargle periodogram occurred for the given series |
| P0 | maximum power in the inspected frequency interval of the Lomb-Scargle periodogram for the given series |
| <axis> | Definition |
|---|---|
| X | X - axis (accelerometer) |
| Y | Y - axis (accelerometer) |
| Z | Z - axis (accelerometer) |
| AA | Average Acceleration |
| AJ | Average Jerk |
Where,
AA = sqrt(X^2 + Y^2 + Z^2)
AJ = sqrt((diff(X))^2 + (diff(Y))^2 + (diff(Z))^2)
There are three more features for walk that don't fit the above category:
| Feature Name | Definition |
|---|---|
| corXY | correlation between the X and Y acceleration |
| corXZ | correlation between the X and Z acceleration |
| corYZ | correlation between the Y and Z acceleration |
The following features used were extracted using mpowertools [https://github.com/itismeghasyam/mpowertools].
| Feature Name | Definition |
|---|---|
| meanAA | mean of the average acceleration series |
| sdAA | standard deviation of the average acceleration series |
| modeAA | mode of the average acceleration series |
| skewAA | skewness of the average acceleration series |
| kurAA | kurtosis of the average acceleration series |
| q1AA | first quartile of the average acceleration series |
| medianAA | median of the average acceleration series |
| q3AA | third quartile of the average acceleration series |
| iqrAA | interquartile range of the average acceleration series |
| rangeAA | range of the average acceleration series |
| acfAA | autocorrelation (lag = 1) of the average acceleration series |
| zcrAA | zero-crossing rate of the average acceleration series |
| dfaAA | scaling exponent of the detrended fluctuation analysis of the average acceleration series |
| turningTime | turning time |
| Postpeak | posture peak |
| Postpower | posture power |
| Alpha | scaling exponent of the detrended fluctuation analysis of the force vector magnitute series |
| dVol | displacement volume (volume of the box around the displacement across the X, Y, and Z directions) |
| ddVol | delta displacement volume |
The following features used were extracted using mhealthtools [https://github.com/Sage-Bionetworks/mhealthtools].
The features are of the format <feature>.<domain>.<IMF>.<stat>_<axis>_<sensorType>.[STAT]
For example skewness.tm.IMF1.md_uav_gyroscope.IQR
| <feature> | Definition |
|---|---|
| cent | centroid |
| complexity | sqrt(var(diff(diff(values)*sampling_rate)*sampling_rate) /var(diff(values)*sampling_rate))) |
| dfa | scaling exponent of the detrended fluctuation analysis |
| energy | energy (or sum of squares of values) |
| EnergyBandN | energy in frequency band N |
| ewt.permEnt [1] | Permutation Entropy of the energies of signals obtained using Empirical Wavelet Transform |
| ewt.shannonEnt [1] | Permutation Entropy of the energies of signals obtained using Empirical Wavelet Transform |
| ewt.simpsonEnt [1] | Permutation Entropy of the energies of signals obtained using Empirical Wavelet Transform |
| ewt.renyiEnt [1] | Permutation Entropy of the energies of signals obtained using Empirical Wavelet Transform |
| ewt.tsallisEnt [1] | Permutation Entropy of the energies of signals obtained using Empirical Wavelet Transform |
| IQR | Inter quantile range |
| kurt | kurtosis |
| kurtosis | kurtosis |
| md | median |
| mean | mean |
| mn*(for time) | min |
| mn*(for frequency) | mean |
| median | median |
| mobility | sqrt(var(diff(values)*sampling_rate)/var(values)) |
| mode | mode, most occuring value |
| mod | most prominent frequency (max magnitude in frequency spectrum) |
| mtkeo | mean of result of teager-kaiser energy operator on the given series |
| mx | max |
| Q25 | first Quartile cutoff |
| Q75 | third Quartile cutoff |
| range | maximum - minimum |
| rmsmag | Root mean square magnitude |
| rough [2] | roughness or total curvature |
| rugo [3] | rugosity |
| sd | standard deviation |
| sem | standard deviation per frequency bin of the spectrum |
| sfm | spectral flatness measure |
| sh | Shannon entropy |
| skew | skewness |
| skewness | skewness |
[1] - https://github.com/Sage-Bionetworks/mhealthtools/blob/master/R/feature_extraction_functions.R#L107
[2] - https://rdrr.io/cran/seewave/man/roughness.html
[3] - https://rdrr.io/cran/seewave/man/rugo.html
| <domain> | Definition |
|---|---|
| tm | time |
| fr | frequency (FFT) |
| <IMF> | Definition |
|---|---|
| IMFN [4] | Nth Intrinsic Mode Function obtained from Emprirical Mode Decomposition |
[4] - https://www.rdocumentation.org/packages/EMD/versions/1.5.8/topics/emd
| <stat> | Definition |
|---|---|
| md | median |
| iqr | interquantile range |
Scope of <stat> is at record level. ie., for eg., skewness.tm.IMF1.md_uav_gyroscope is obtained
by taking the median of the feature obtained from skewnwss, tm, IMF1, uav, gyrscope across all windows for a given record
| <axis> | Definition |
|---|---|
| ua | user linear acceleration (accelerometer) |
| ud | user linear displacement (accelerometer) |
| uv | user linear velocity (accelerometer) |
| uj | user linear jerk (accelerometer) |
| uaacf | autocorrelation function of user acceleration (accelerometer) |
| uad | user rotational displacement (gyroscope) |
| uav | user rotational velocity (gyroscope) |
| uavacf | autocorrelation function of user rotational velocity (gyroscope) |
| uaa | user rotational acceleration (gyroscope) |
ua is the gravity corrected raw accelerometer values in m/s^2
uav is the raw rotational velocity collected in rad/s by the gyroscope
| <sensorType> | Definition |
|---|---|
| acceleromter | accelerometer |
| gyroscope | gyroscope |
| [STAT] | Definition |
|---|---|
| md | median |
| iqr | interquantile range |
[STAT] is optional. Scope of [STAT] is at individual level. ie., for eg., skewness.tm.IMF1.md_uav_gyroscope.IQR is obtained
by taking the inter quantile range across all records for a given subject