def _addSyllableNucleiToTextgrids(wavPath, tgPath, tierName,
syllableNucleiPath, outputPath):
# Add syllable nuclei to textgrids
for name in utils.findFiles(wavPath, filterExt=".wav", stripExt=True):
tg = tgio.openTextgrid(join(tgPath, name + ".TextGrid"))
entryList = tg.tierDict[tierName].entryList
startTimeList = [entry[0] for entry in entryList]
nucleusSyllableList = uwe_sr.toAbsoluteTime(name, syllableNucleiPath,
startTimeList)
flattenedSyllableList = [nuclei for sublist in nucleusSyllableList
for nuclei in sublist]
wavFN = join(wavPath, name + ".wav")
duration = audio_scripts.getSoundFileDuration(wavFN)
oom = my_math.orderOfMagnitude(len(flattenedSyllableList))
labelTemplate = "%%0%dd" % (oom + 1)
entryList = [(timestamp, labelTemplate % i)
for i, timestamp in enumerate(flattenedSyllableList)]
print(flattenedSyllableList)
tier = tgio.PointTier("Syllable Nuclei", entryList, 0, duration)
tgFN = join(tgPath, name + ".TextGrid")
tg = tgio.openTextgrid(tgFN)
tg.addTier(tier)
tg.save(join(outputPath, name + ".TextGrid"))
def detectPitchErrors(pitchList, maxJumpThreshold=0.70, tgToMark=None):
'''
Detect pitch halving and doubling errors.
If a textgrid is passed in, it adds the markings to the textgrid
'''
assert (maxJumpThreshold >= 0.0 and maxJumpThreshold <= 1.0)
errorList = []
for i in range(1, len(pitchList)):
lastPitch = pitchList[i - 1][1]
currentPitch = pitchList[i][1]
ceilingCutoff = currentPitch / maxJumpThreshold
floorCutoff = currentPitch * maxJumpThreshold
if ((lastPitch <= floorCutoff) or (lastPitch >= ceilingCutoff)):
currentTime = pitchList[i][0]
errorList.append([currentTime, currentPitch / lastPitch])
if tgToMark is not None:
tierName = "pitch errors"
assert (tierName not in tgToMark.tierNameList)
pointTier = tgio.PointTier(tierName, errorList, tgToMark.minTimestamp,
tgToMark.maxTimestamp)
tgToMark.addTier(pointTier)
return errorList, tgToMark
def make_textgrid(df, out_name, orig_name=None, word2phone=None):
if orig_name:
tg = tgio.openTextgrid(orig_name)
else:
tg = tgio.Textgrid()
phones_list = []
syllables_list = []
curr_syllable = []
for tup in df[['start', 'end', 'phone']].itertuples():
phones_list.append((tup.start, tup.end, tup.phone))
if tup.phone in set(['spn', 'sil']):
# pass
syllables_list.append((tup.start, tup.end, tup.phone))
curr_syllable = []
elif len(tup.phone) > 2 and tup.phone[-2] == '_': # final
curr_syllable.append(tup.phone)
syllables_list.append(
(initial_start, tup.end, ' '.join(curr_syllable)))
curr_syllable = []
else: # initial
curr_syllable.append(tup.phone)
initial_start = tup.start
phone_tier = tgio.IntervalTier('phone', phones_list)
syllable_tier = tgio.IntervalTier('syllable\_phones', syllables_list)
if orig_name and word2phone:
ipus, xmins, xmaxs = get_ipus(tg)
word_list, unmatched_words, break_list = make_word_list(
syllable_tier, ipus, word2phone, out_name, xmaxs)
word_tier = tgio.IntervalTier('word', word_list)
tg.addTier(word_tier)
tg.addTier(phone_tier)
tg.addTier(syllable_tier)
if not tg.tierDict['breaks'].entryList:
tg.removeTier('breaks')
break_tier = tgio.PointTier('break', break_list)
tg.addTier(break_tier)
else:
print(out_name, 'has break tier, did not write new one')
os.makedirs(os.path.dirname(out_name), exist_ok=True)
tg.save(out_name, useShortForm=False)
print('wrote to {}, # matched: {}, # unmatched: {}'.format(
out_name, len(word_list), len(unmatched_words)))
return len(word_list), len(unmatched_words)