def extractAxisFeature(dfData,
strDataName,
strCol,
dSamplingFreq,
nReponses,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration,
nFreqLowCut=20,
nFreqHighCut=120):
"""
Extract feature for single-axis data as follows:
1. divide this data into multiple responses;
2. divide each response into several segments;
3. extract features from all segment;
4. combine all features together;
Parameters:
----
dfData:
the data frame
strDataName:
name of data
strCol:
axis name
dSamplingFreq:
sampling frequency of data
nResponse:
number of responses in single-axis data
nSegmentPerRespsonse:
number of excitation segment per response
dVibrationDuration:
duration of vibration (seconds)
dIntervalDuration:
interval duration btw segments (seconds)
dRestDuration:
the static duration btw responses
nFreqLowCut:
the low cutting frequency of bandpass filter
nFreqHighCut:
the high cutting frequency of bandpass filter
Returns:
----
dcResponseFeature:
a dict of key-value pairs like:
{'r1': {
'x0_s1_std': 0.85,
'y0_s5_min': 4.72,
...
}
}
"""
arrData = dfData[strCol].values
# bandpass filter to select spectrum of interest
arrFiltered = bp_filter.butter_bandpass_filter(arrData,
lowcut=nFreqLowCut,
highcut=nFreqHighCut,
fs=dSamplingFreq,
order=7)
# find reponses & segements in each response
lsResponses, arrResponseEndIndex, \
arrLocalPeakRefVal = sd.splitData(arrFiltered,
dSamplingFreq,
nReponses,
nSegmentPerRespsonse,
dVibrationDuration,
dIntervalDuration,
dRestDuration)
dcResponseFeature = {}
for nRespID, lsSegments in enumerate(lsResponses):
dcFeaturePerResponse = dcResponseFeature.get( \
"%s_r%d"% (strDataName, nRespID), None)
if(dcFeaturePerResponse == None):
dcFeaturePerResponse = {}
dcResponseFeature["%s_r%d"% (strDataName, nRespID)] = \
dcFeaturePerResponse
for nSegID, (nSegStart, nSegEnd) in enumerate(lsSegments):
arrSeg = arrFiltered[nSegStart:nSegEnd]
dcSegFeatures = extractSegmentFeatures(strCol, nSegID, arrSeg)
dcFeaturePerResponse.update(dcSegFeatures)
return dcResponseFeature
bPlotRawData = True
bPlotSegmentLine = True
bPlotModulus = True
lsColors = lsRGB*60
# lsColors = ['#65737e']*3
nRows= len(lsAxis2Inspect)+1 if bPlotModulus else len(lsAxis2Inspect)
nCols = 2 if bPlotRawData is True else 1
fig, axes = plt.subplots(nrows=nRows, ncols=nCols, squeeze=False)
lsResponses, arrResponseEndIndex, arrBandwidthSTD = None, None, None
for i, col in enumerate(lsAxis2Inspect):
arrData = dfData[col].values
arrFiltered = bp_filter.butter_bandpass_filter(arrData, lowcut=20,
highcut=130,
fs=dSamplingFreq,
order=10)
axes[i%nRows, 0].plot(arrFiltered, color=lsColors[i])
axes[i%nRows, 0].set_xlabel("%s_filtered" % col)
# axes[i%nRows, 0].set_ylim(-700, 700)
# plot raw data
if(bPlotRawData is True):
axes[i%nRows, 1].plot(arrData, color=lsColors[i])
axes[i%nRows, 1].set_xlabel(col)
# plot segement lines
if(bPlotSegmentLine is True):
lsResponses, arrResponseEndIndex, \
arrBandwidthSTD = splitData(arrFiltered,
nRows = len(lsAxis2Inspect)
nCols = len(lsFileNames)
nFigWidth, nFigHeight = plt.figaspect(4.0/3.0)
fig, axes = plt.subplots(nrows=nRows, ncols=nCols,
squeeze=False, sharex=True,
sharey=True,
figsize=(nFigWidth, nFigHeight) )
for nDataID, strFileName in enumerate(lsFileNames):
dfData = sd.loadData(strWorkingDir, strFileName, lsColumnNames)
for nColID, strCol in enumerate(lsAxis2Inspect):
arrData = dfData[strCol].values
arrFiltered = bp_filter.butter_bandpass_filter(arrData,
lowcut=20,
highcut=120,
fs=dSamplingFreq,
order=10)
axes[nColID, nDataID].plot(arrFiltered, color=lsColors[nColID])
axes[nColID, nDataID].set_ylim(-5000, 5000)
axes[nColID, nDataID].set_ylabel("%s" % strCol,
rotation=0,
fontname=strBasicFontName,
fontsize=nBasicFontSize)
axes[nColID, nDataID].set_xlabel("Time(sec)",
rotation=0,
fontname=strBasicFontName,
fontsize=nBasicFontSize)
xTickLabels = axes[nColID, nDataID].xaxis.get_ticklabels()
plt.setp(xTickLabels, fontname=strBasicFontName,
size=nBasicFontSize)
