132 lines
6.2 KiB
Python
132 lines
6.2 KiB
Python
import numpy as np
|
|
import pandas as pd
|
|
|
|
|
|
def convertInputInto2d(input, windowLength, overlap=0):
|
|
"""Convert input into 2d matrix with width = numCols. The last row is padded with zeros to match the other rows.
|
|
|
|
Overlap has to be smaller than window length
|
|
|
|
:param input: the one dimensional array
|
|
:param windowLength: window length, expressed in number of samples
|
|
:param overlap: Amount of overlap
|
|
:return: 2D matrix
|
|
"""
|
|
|
|
if windowLength <= overlap:
|
|
raise Exception("Overlap has to be smaller than window length")
|
|
|
|
inputWasList = True
|
|
if type(input) != list:
|
|
inputWasList = False
|
|
input = input.tolist()
|
|
|
|
out = [input[i: i + windowLength] for i in range(0, len(input), windowLength - overlap)]
|
|
out[-1].extend([0] * (windowLength - len(out[-1])))
|
|
return out if inputWasList else np.asarray(out)
|
|
|
|
|
|
def convertInputInto2dTime(input, timeThreshold):
|
|
""" Convert input array into 2D matrix by time interval. When the timeThreshold is reached in each row,
|
|
the process continues in the next row.
|
|
|
|
:param input: the pandas dataframe with rows "time" and "data"
|
|
:param timeThreshold: the threshold with which the row width is defined
|
|
:return: 2D matrix
|
|
"""
|
|
outData = [[]]
|
|
outTime = [[]]
|
|
startTime = 0
|
|
|
|
for index, row in input.iterrows():
|
|
t = row["time"]
|
|
data = row["data"]
|
|
|
|
if t - startTime >= timeThreshold:
|
|
startTime = t
|
|
outData.append([])
|
|
outTime.append([])
|
|
|
|
outData[-1].append(data)
|
|
outTime[-1].append(t)
|
|
|
|
return outData, outTime
|
|
|
|
|
|
def convert1DEmpaticaToArray(pathToEmpaticaCsvFile):
|
|
""" Convert 1D empatica file to array
|
|
|
|
:param pathToEmpaticaCsvFile: path to Empatica csv file
|
|
:return: array of data, starting timestamp of data, sample rate of data
|
|
"""
|
|
df = pd.read_csv(pathToEmpaticaCsvFile, names=["name"])
|
|
startTimeStamp = df.name[0]
|
|
sampleRate = df.name[1]
|
|
df.drop([0, 1], inplace=True)
|
|
data = df.name.ravel()
|
|
return data, startTimeStamp, sampleRate
|
|
|
|
|
|
def convert3DEmpaticaToArray(pathToEmpaticaCsvFile):
|
|
""" Convert 3D empatica file to array
|
|
|
|
:param pathToEmpaticaCsvFile: path to Empatica csv file
|
|
:return: array of data, starting timestamp of data, sample rate of data
|
|
"""
|
|
df = pd.read_csv(pathToEmpaticaCsvFile, names=["x", "y", "z"])
|
|
startTimeStamp = df.x[0]
|
|
sampleRate = df.x[1]
|
|
df.drop([0, 1], inplace=True)
|
|
data = np.vstack((df.x.ravel(), df.y.ravel(), df.z.ravel()))
|
|
return data, startTimeStamp, sampleRate
|
|
|
|
|
|
def checkForFeature(featureName, featureNames):
|
|
return featureNames is None or featureName in featureNames
|
|
|
|
|
|
frequencyFeatureNames = ["fqHighestPeakFreqs", "fqHighestPeaks", "fqEnergyFeat", "fqEntropyFeat", "fqHistogramBins",
|
|
"fqAbsMean", "fqSkewness", "fqKurtosis", "fqInterquart"]
|
|
|
|
genericFeatureNames = ["autocorrelations", "countAboveMean", "countBelowMean", "maximum", "minimum", "meanAbsChange",
|
|
"longestStrikeAboveMean", "longestStrikeBelowMean", "stdDev", "median", "meanChange",
|
|
"numberOfZeroCrossings", "absEnergy", "linearTrendSlope", "ratioBeyondRSigma", "binnedEntropy",
|
|
"numOfPeaksAutocorr", "numberOfZeroCrossingsAutocorr", "areaAutocorr",
|
|
"calcMeanCrossingRateAutocorr", "countAboveMeanAutocorr", "sumPer", "sumSquared",
|
|
"squareSumOfComponent", "sumOfSquareComponents"]
|
|
|
|
accelerometerFeatureNames = ["meanLow", "areaLow", "totalAbsoluteAreaBand", "totalMagnitudeBand", "entropyBand",
|
|
"skewnessBand", "kurtosisBand", "postureDistanceLow", "absoluteMeanBand",
|
|
"absoluteAreaBand", "quartilesBand", "interQuartileRangeBand",
|
|
"varianceBand", "coefficientOfVariationBand", "amplitudeBand", "totalEnergyBand",
|
|
"dominantFrequencyEnergyBand", "meanCrossingRateBand", "correlationBand",
|
|
"quartilesMagnitudesBand",
|
|
"interQuartileRangeMagnitudesBand", "areaUnderAccelerationMagnitude", "peaksDataLow",
|
|
"sumPerComponentBand", "velocityBand", "meanKineticEnergyBand",
|
|
"totalKineticEnergyBand", "squareSumOfComponent", "sumOfSquareComponents",
|
|
"averageVectorLength", "averageVectorLengthPower", "rollAvgLow", "pitchAvgLow",
|
|
"rollStdDevLow", "pitchStdDevLow",
|
|
"rollMotionAmountLow", "rollMotionRegularityLow", "manipulationLow", "rollPeaks",
|
|
"pitchPeaks",
|
|
"rollPitchCorrelation"]
|
|
|
|
gyroscopeFeatureNames = ["meanLow", "areaLow", "totalAbsoluteAreaLow", "totalMagnitudeLow", "entropyLow", "skewnessLow",
|
|
"kurtosisLow",
|
|
"quartilesLow", "interQuartileRangeLow", "varianceLow", "coefficientOfVariationLow",
|
|
"amplitudeLow",
|
|
"totalEnergyLow", "dominantFrequencyEnergyLow", "meanCrossingRateLow", "correlationLow",
|
|
"quartilesMagnitudeLow", "interQuartileRangeMagnitudesLow", "areaUnderMagnitude",
|
|
"peaksCountLow",
|
|
"averageVectorLengthLow", "averageVectorLengthPowerLow"]
|
|
|
|
gsrFeatureNames = ['mean', 'std', 'q25', 'q75', 'qd', 'deriv', 'power', 'numPeaks', 'ratePeaks', 'powerPeaks',
|
|
'sumPosDeriv', 'propPosDeriv', 'derivTonic', 'sigTonicDifference', 'freqFeats',
|
|
'maxPeakAmplitudeChangeBefore', 'maxPeakAmplitudeChangeAfter',
|
|
'avgPeakAmplitudeChangeBefore', 'avgPeakAmplitudeChangeAfter', 'avgPeakChangeRatio',
|
|
'maxPeakIncreaseTime', 'maxPeakDecreaseTime', 'maxPeakDuration', 'maxPeakChangeRatio',
|
|
'avgPeakIncreaseTime', 'avgPeakDecreaseTime', 'avgPeakDuration', 'maxPeakResponseSlopeBefore',
|
|
'maxPeakResponseSlopeAfter', 'signalOverallChange', 'changeDuration', 'changeRate',
|
|
'significantIncrease', 'significantDecrease']
|
|
|
|
hrvFeatureNames = ['meanHr', 'ibi', 'sdnn', 'sdsd', 'rmssd', 'pnn20', 'pnn50', 'sd', 'sd2', 'sd1/sd2', 'numRR']
|