def extractPraatPitchForEpochs(pitchPath, epochPath, tgInfoPath, outputPath):
utils.makeDir(outputPath)
for fn in utils.findFiles(pitchPath, filterExt=".txt"):
name = os.path.splitext(fn)[0]
print name
epochList = utils.openCSV(epochPath, fn)
epochList = [(epochNum, float(start), float(stop)) for epochNum, start, stop in epochList]
entryList = utils.openCSV(tgInfoPath, fn)
entryList = [(float(start), float(stop), label) for start, stop, label in entryList]
dataList = praat_pi.loadPitchAndTime(pitchPath, fn)
# Get F0 values for the intervals when the mother was speaking
speechDataList = []
for start, stop, label in entryList:
speechDataList.extend(praat_pi.getAllValuesInTime(start, stop, dataList))
# Get F0 values for the times the mother is speaking for each epoch
pitchData = []
for epochNum, start, stop in epochList:
start, stop = float(start), float(stop)
duration = stop - start
epochValueList = praat_pi.getAllValuesInTime(start, stop, speechDataList)
f0List = [f0Val for time, f0Val, intVal in epochValueList]
pitchData.append(praat_pi.extractPitchMeasuresForSegment(f0List, name, epochNum, medianFilterWindowSize=None, filterZeroFlag=True))
open(join(outputPath, "%s.txt" % name), "w").write("\n".join(pitchData) + "\n")
def aggregateSpeechRate(tgInfoPath, speechRatePath, outputPath, samplingRate):
utils.makeDir(outputPath)
finishedList = utils.findFiles(outputPath, filterExt=".txt")
for fn in utils.findFiles(tgInfoPath, filterExt=".txt",
skipIfNameInList=finishedList):
# Load subset speech rate
name = os.path.splitext(fn)[0]
speechRateFNList = utils.findFiles(speechRatePath, filterExt=".txt",
filterPattern=name)
subSplitList = utils.openCSV(tgInfoPath, fn)
# Convert the sample numbers to seconds
# They are in terms of the beginning of the subset they are in but
# need to be in terms of the start of the file the larger file the
# subset originated from
outputList = []
for splitInfo, speechRateFN in utils.safeZip([subSplitList,
speechRateFNList],
enforceLength=True):
start, stop, label = splitInfo
speechRateList = utils.openCSV(speechRatePath, speechRateFN, valueIndex=0)
speechRateList = [value for value in speechRateList if value != '']
speechRateList = [str(float(start) + float(sampleNum) / float(samplingRate)) for sampleNum in speechRateList]
outputList.append( ",".join(speechRateList) )
open(join(outputPath, fn), "w").write("\n".join(outputList) + "\n")
def manualPhoneCount(tgInfoPath, isleFN, outputPath, skipList=None):
if skipList is None:
skipList = []
utils.makeDir(outputPath)
isleDict = isletool.LexicalTool(isleFN)
existFNList = utils.findFiles(outputPath, filterPaths=".txt")
for fn in utils.findFiles(tgInfoPath, filterExt=".txt",
skipIfNameInList=existFNList):
if os.path.exists(join(outputPath, fn)):
continue
print(fn)
dataList = utils.openCSV(tgInfoPath, fn)
dataList = [row[2] for row in dataList] # start, stop, tmpLabel
outputList = []
for tmpLabel in dataList:
if tmpLabel not in skipList:
syllableCount, phoneCount = isletool.getNumPhones(isleDict,
tmpLabel,
maxFlag=True)
else:
syllableCount, phoneCount = 0, 0
outputList.append("%d,%d" % (syllableCount, phoneCount))
outputTxt = "\n".join(outputList)
with open(join(outputPath, fn), "w") as fd:
fd.write(outputTxt)
def toAbsoluteTime(namePrefix, matlabOutputPath, startTimeList):
'''
Converts the sampled times from relative to absolute time
The input may be split across a number of files. This script assumes
that files of the pattern <<namePrefix>><<nameSuffix>>.txt correspond
to different parts of the same source file.
namePrefix - name of the original wav file with no suffix
speechRatePath - the path where the output of the matlab script is placed
startTimeList - there needs to be one file here for each file in
speechRatePath with the pattern namePrefix
Returns a list of lists where each sublist corresponds to the output of
one file matching <<namePrefix>>
'''
# Load subset speech rate
speechRateFNList = utils.findFiles(matlabOutputPath,
filterExt=".txt",
filterPattern=namePrefix)
returnList = []
for start, speechRateFN in utils.safeZip([startTimeList, speechRateFNList],
enforceLength=True):
speechRateList = utils.openCSV(matlabOutputPath,
speechRateFN,
valueIndex=0)
speechRateList = [value for value in speechRateList if value != '']
speechRateList = [
str(float(start) + float(sampNum)) for sampNum in speechRateList
]
returnList.append(speechRateList)
return returnList
def toAbsoluteTime(namePrefix, matlabOutputPath, startTimeList):
'''
Converts the sampled times from relative to absolute time
The input may be split across a number of files. This script assumes
that files of the pattern <<namePrefix>><<nameSuffix>>.txt correspond
to different parts of the same source file.
namePrefix - name of the original wav file with no suffix
speechRatePath - the path where the output of the matlab script is placed
startTimeList - there needs to be one file here for each file in
speechRatePath with the pattern namePrefix
Returns a list of lists where each sublist corresponds to the output of
one file matching <<namePrefix>>
'''
# Load subset speech rate
speechRateFNList = utils.findFiles(matlabOutputPath, filterExt=".txt",
filterPattern=namePrefix)
returnList = []
for start, speechRateFN in utils.safeZip([startTimeList, speechRateFNList],
enforceLength=True):
speechRateList = utils.openCSV(matlabOutputPath,
speechRateFN,
valueIndex=0)
speechRateList = [value for value in speechRateList if value != '']
speechRateList = [str(float(start) + float(sampNum))
for sampNum in speechRateList]
returnList.append(speechRateList)
return returnList
def removeFilledPauses(inputPath, outputPath):
utils.makeDir(outputPath)
for fn in utils.findFiles(inputPath, filterExt=".txt"):
dataList = utils.openCSV(inputPath, fn)
dataList = [[start, stop, label] for start, stop, label in dataList if label == "MS"]
dataList = [",".join(row) for row in dataList]
open(join(outputPath, fn), "w").write("\n".join(dataList) + "\n")
def adjustEpochNumbers(inputPath, outputPath):
utils.makeDir(outputPath)
for fn in utils.findFiles(inputPath, filterExt=".txt"):
dataList = utils.openCSV(inputPath, fn)
dataList = ["%02d,%s,%s" % (int(id)+1,start, stop)
for id, start, stop in dataList]
open(join(outputPath, fn), "w").write("\n".join(dataList) + "\n")
def manualPhoneCountForEpochs(manualCountsPath, tgInfoPath, epochPath,
outputPath):
utils.makeDir(outputPath)
skipList = utils.findFiles(outputPath, filterExt=".txt")
for fn in utils.findFiles(tgInfoPath, filterExt=".txt",
skipIfNameInList=skipList):
epochList = utils.openCSV(epochPath, fn)
tgInfo = utils.openCSV(tgInfoPath, fn)
manualCounts = utils.openCSV(manualCountsPath, fn)
epochOutputList = []
for epochTuple in epochList: # Epoch num, start, stop
epochStart, epochStop = float(epochTuple[1]), float(epochTuple[2])
# Find all of the intervals that are at least partially
# contained within the current epoch
epochSyllableCount = 0
epochPhoneCount = 0
speechDuration = 0
for info, counts in utils.safeZip([tgInfo, manualCounts],
enforceLength=True):
start, stop = float(info[0]), float(info[1])
syllableCount, phoneCount = float(counts[0]), float(counts[1])
# Accounts for intervals that straddle an epoch boundary
multiplicationFactor = percentInside(start, stop,
epochStart, epochStop)
speechDuration += (stop - start) * multiplicationFactor
epochSyllableCount += syllableCount * multiplicationFactor
epochPhoneCount += phoneCount * multiplicationFactor
epochOutputList.append("%f,%f,%f" % (epochSyllableCount,
epochPhoneCount,
speechDuration))
with open(join(outputPath, fn), "w") as fd:
fd.write("\n".join(epochOutputList))
def generateEpochRowHeader(epochPath, outputPath, sessionCode):
utils.makeDir(outputPath)
for fn in utils.findFiles(epochPath, filterExt=".txt"):
epochList = utils.openCSV(epochPath, fn)
id = fn.split("_")[2]
outputList = [",".join([id, sessionCode, epoch, epochStart, epochEnd, str(float(epochEnd) - float(epochStart))]) for epoch, epochStart, epochEnd in epochList]
open(join(outputPath, fn), "w").write("\n".join(outputList) + "\n")
def aggregateFeatures(featurePath, featureList, headerStr=None):
outputDir = join(featurePath, "aggr")
utils.makeDir(outputDir)
fnList = []
dataList = []
# Find the files that exist in all features
for feature in featureList:
fnSubList = utils.findFiles(join(featurePath, feature),
filterExt=".txt")
fnList.append(fnSubList)
actualFNList = []
for featureFN in fnList[0]:
if all([featureFN in subList for subList in fnList]):
actualFNList.append(featureFN)
for featureFN in actualFNList:
dataList = []
for feature in featureList:
featureDataList = utils.openCSV(join(featurePath, feature),
featureFN,
encoding="utf-8")
dataList.append([",".join(row) for row in featureDataList])
name = os.path.splitext(featureFN)[0]
dataList.insert(0, [name for _ in range(len(dataList[0]))])
tDataList = utils.safeZip(dataList, enforceLength=True)
outputList = [",".join(row) for row in tDataList]
outputTxt = "\n".join(outputList)
outputFN = join(outputDir, name + ".csv")
with io.open(outputFN, "w", encoding="utf-8") as fd:
fd.write(outputTxt)
# Cat all files together
aggrOutput = []
if headerStr is not None:
aggrOutput.append(headerStr)
for fn in utils.findFiles(outputDir, filterExt=".csv"):
if fn == "all.csv":
continue
with io.open(join(outputDir, fn), "r", encoding='utf-8') as fd:
aggrOutput.append(fd.read())
with io.open(join(outputDir, "all.csv"), "w", encoding='utf-8') as fd:
fd.write("\n".join(aggrOutput))
def aggregateFeatures(featurePath, featureList, headerStr=None):
outputDir = join(featurePath, "aggr")
utils.makeDir(outputDir)
fnList = []
dataList = []
# Find the files that exist in all features
for feature in featureList:
fnSubList = utils.findFiles(join(featurePath, feature),
filterExt=".txt")
fnList.append(fnSubList)
actualFNList = []
for featureFN in fnList[0]:
if all([featureFN in subList for subList in fnList]):
actualFNList.append(featureFN)
for featureFN in actualFNList:
dataList = []
for feature in featureList:
featureDataList = utils.openCSV(join(featurePath, feature),
featureFN, encoding="utf-8")
dataList.append([",".join(row) for row in featureDataList])
name = os.path.splitext(featureFN)[0]
dataList.insert(0, [name for _ in range(len(dataList[0]))])
tDataList = utils.safeZip(dataList, enforceLength=True)
outputList = [",".join(row) for row in tDataList]
outputTxt = "\n".join(outputList)
outputFN = join(outputDir, name + ".csv")
with io.open(outputFN, "w", encoding="utf-8") as fd:
fd.write(outputTxt)
# Cat all files together
aggrOutput = []
if headerStr is not None:
aggrOutput.append(headerStr)
for fn in utils.findFiles(outputDir, filterExt=".csv"):
if fn == "all.csv":
continue
with io.open(join(outputDir, fn), "r", encoding='utf-8') as fd:
aggrOutput.append(fd.read())
with io.open(join(outputDir, "all.csv"), "w", encoding='utf-8') as fd:
fd.write("\n".join(aggrOutput))
def uwePhoneCountForEpochs(epochPath, tgInfoPath, manualCountsPath, outputPath):
utils.makeDir(outputPath)
for fn in utils.findFiles(tgInfoPath, filterExt=".txt"):
print fn
epochList = utils.openCSV(epochPath, fn)
tgInfo = utils.openCSV(tgInfoPath, fn)
manualCounts = utils.openCSV(manualCountsPath, fn)
epochOutputList = []
for epochNumber, epochStart, epochStop in epochList:
epochStart, epochStop = float(epochStart), float(epochStop)
# Find all of the intervals that are at least partially contained within
# the current epoch
epochSyllableCount = 0
unadjustedEpochSyllableCount = 0
epochArticulationRate = 0
epochAverageSyllableDuration = 0
for info, nucleusList in utils.safeZip([tgInfo, manualCounts],
enforceLength=True):
start, stop, wordList = info
start, stop = float(start), float(stop)
syllableCount = len(nucleusList)
unadjustedEpochSyllableCount += syllableCount
# Accounts for intervals that straddle an epoch boundary
multiplicationFactor = _percentInside(start, stop, epochStart,
epochStop)
epochSyllableCount += syllableCount * multiplicationFactor
# epochOutputList.append("%f,%f" % (unadjustedEpochSyllableCount,epochSyllableCount))
epochOutputList.append("%f" % (epochSyllableCount))
open(join(outputPath, fn), "w").write("\n".join(epochOutputList) + "\n")
def addEpochsToTextgrids(tgPath, epochPath, outputPath):
utils.makeDir(outputPath)
for name in utils.findFiles(tgPath, filterExt=".TextGrid", stripExt=True):
print name
tg = tgio.openTextGrid(join(tgPath, name+".TextGrid"))
entryList = utils.openCSV(epochPath, name+".txt")
entryList = [(float(start), float(end), label) for label, start, end in entryList]
tier = tgio.IntervalTier("epochs", entryList, minT=0, maxT=tg.maxTimestamp)
tg.addTier(tier)
tg.save(join(outputPath, name+".TextGrid"))
def findFrequenciesForWordLists(featurePath, countObj, frequencyNormFunc):
frequencyPath = join(featurePath, "frequency")
utils.makeDir(frequencyPath)
wordsPath = join(featurePath, "words")
for fn in utils.findFiles(wordsPath):
wordList = utils.openCSV(wordsPath, fn, valueIndex=0, encoding="utf-8")
countList = []
for word in wordList:
tmp = countObj.getFrequency(word,
frequencyNormFunc,
outOfDictionaryValue=1)
count, freq, logFreq = tmp
countList.append("%f,%f,%f" % (count, freq, logFreq))
with open(join(frequencyPath, fn), "w") as fd:
fd.write("\n".join(countList))
def findFrequenciesForWordLists(featurePath, countObj, frequencyNormFunc):
frequencyPath = join(featurePath, "frequency")
utils.makeDir(frequencyPath)
wordsPath = join(featurePath, "words")
for fn in utils.findFiles(wordsPath):
wordList = utils.openCSV(wordsPath, fn, valueIndex=0, encoding="utf-8")
countList = []
for word in wordList:
tmp = countObj.getFrequency(word,
frequencyNormFunc,
outOfDictionaryValue=1)
count, freq, logFreq = tmp
countList.append("%f,%f,%f" % (count, freq, logFreq))
with open(join(frequencyPath, fn), "w") as fd:
fd.write("\n".join(countList))
def medianFilter(f0Path, outputPath, windowSize):
# windowSize must be odd
assert (windowSize % 2 != 0)
utils.makeDir(outputPath)
for fn in utils.findFiles(f0Path, filterExt=".txt"):
valueList = utils.openCSV(f0Path, fn)
f0List = [
float(row[1]) if row[1] != "--undefined--" else 0
for row in valueList
] # time, f0Val, intensityVal
f0Filtered = filters.medianFilter(f0List,
windowSize,
useEdgePadding=True)
outputList = [
"%s,%0.3f,%s" % (row[0], f0Val, row[2])
for row, f0Val in zip(*[valueList, f0Filtered])
]
open(join(outputPath, fn), "w").write("\n".join(outputList) + "\n")
def eventStructurePerEpoch(epochPath, fullyFilteredTGPath,
childFilteredTGPath, noiseFilteredTGPath,
unfilteredTGPath, outputPath,
speechTierName, laughterTierName):
'''
How frequent and with what duration did laughter, pauses, and speech occur
'''
def _getCountsAndDurations(tier, searchLabel):
entryList = tier.find(searchLabel)
durationList = [float(stop) - float(start)
for start, stop, label in entryList]
count = len(entryList)
return sum(durationList), count
utils.makeDir(outputPath)
for name in utils.findFiles(epochPath, filterExt=".txt", stripExt=True):
epochList = utils.openCSV(epochPath, name+".txt")
epochList = [(epochNum, float(start), float(stop))
for epochNum, start, stop in epochList]
tg = tgio.openTextGrid(join(fullyFilteredTGPath,
name + ".TextGrid"))
childFilteredTG = tgio.openTextGrid(join(childFilteredTGPath,
name + ".TextGrid"))
noiseFilteredTG = tgio.openTextGrid(join(noiseFilteredTGPath,
name + ".TextGrid"))
origTG = tgio.openTextGrid(join(unfilteredTGPath,
name + ".TextGrid"))
outputList = []
for epochNum, start, stop in epochList:
subTG = tg.crop(strictFlag=False, softFlag=False,
startTime=start, endTime=stop)
speechTier = subTG.tierDict[speechTierName]
laughterTier = subTG.tierDict[laughterTierName]
pauseDur, numPauses = _getCountsAndDurations(speechTier, "FP")
speechDur, numSpeech = _getCountsAndDurations(speechTier, "MS")
laughDur, numLaughter = _getCountsAndDurations(laughterTier, "LA")
subCSFilteredTG = childFilteredTG.crop(strictFlag=False,
softFlag=False,
startTime=start,
endTime=stop)
csFilteredTier = subCSFilteredTG.tierDict[speechTierName]
csFiltSpeech, numCSFiltSpeech = _getCountsAndDurations(csFilteredTier,
"MS")
subNoiseFilteredTG = noiseFilteredTG.crop(strictFlag=False,
softFlag=False,
startTime=start,
endTime=stop)
nsFilteredTier = subNoiseFilteredTG.tierDict[speechTierName]
nsFiltSpeech, numNsFiltSpeech = _getCountsAndDurations(nsFilteredTier,
"MS")
subOrigTG = origTG.crop(strictFlag=False,
softFlag=False,
startTime=start,
endTime=stop)
origSpeechTier = subOrigTG.tierDict[speechTierName]
fullSpeechDur, fullNumSpeech = _getCountsAndDurations(origSpeechTier,
"MS")
epochTuple = (speechDur, numSpeech, csFiltSpeech, nsFiltSpeech,
fullSpeechDur, fullSpeechDur - speechDur,
pauseDur, numPauses, laughDur, numLaughter)
outputList.append("%.02f, %d, %.02f, %.02f, %.02f, %.02f, %.02f, %d, %.02f, %d" % epochTuple)
open(join(outputPath, name+".txt"), "w").write("\n".join(outputList) + "\n")
