# -*- coding: utf-8 -*-
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import sys, os
import copy
from glob import glob
import pylab as pl
#sometimes the limit needs to be increased to pickle large utts...
sys.setrecursionlimit(10000) #default is generally 1000
import ttslab
from ttslab.hrg import Utterance
ttslab.extend(Utterance, "ufuncs_analysis")
import speechlabels as sl
UTTDIR = "build/utts"
UTTDIR2 = "build/qc_utts"
RECDIR = "build/halign/labels"
def parse_logl_from_recs(u, ul, phoneset, absl=False):
frameperiod = 50000
closure_phone = phoneset.features["closure_phone"]
phmap = phoneset.map
assert u["file_id"] == ul.name
"""
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import shutil, os
from tempfile import mkdtemp
from collections import OrderedDict
from copy import deepcopy
import numpy as np
import ttslab
from ttslab.trackfile import Track
ttslab.extend(Track, "tfuncs_analysis")
import speechlabels as sl
NONE_WORD = "NONE"
WAV_EXT = "wav"
FEAT_EXT = "featvecs"
TEXTGRID_EXT = "TextGrid"
SIG2FV = "sig2fv -coefs melcep -delta melcep -melcep_order 12 -fbank_order 24 -shift %(shift)s -factor 5.0 -preemph 0.97 -otype est %(inputfile)s -o %(outputfile)s"
def fill_startendtimes(utt):
""" Use 'end' time feature in segments to fill info for other
items in other common relations..
"""
#segments (are contiguous in time)...
import os import sys from collections import defaultdict from glob import glob import copy from tempfile import mkstemp from ConfigParser import ConfigParser import numpy as np from wav2psmfcc import PMExtractor from make_f0_praat_script import script_writer as F0_PSCWriter from make_f0_praat import f0filler as F0Filler import ttslab import ttslab.hrg as hrg ttslab.extend(hrg.Utterance, "ufuncs_analysis") from ttslab.trackfile import Track SAVE_COMPLETE_UTTS = True #sometimes the limit needs to be increased to pickle large utts... BIGGER_RECURSION_LIMIT = 20000 #default is generally 1000 WAV_DIR = "wavs" PM_DIR = "pm" LPC_DIR = "lpc" F0_DIR = "f0" MCEP_DIR = "mcep" JOIN_DIR = "joincoef" UTT_DIR = "utts" COMPLETE_UTT_DIR = "complete_utts"
import networkx as nx
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt, mpld3
import numpy as np
from tempfile import mkdtemp
try:
from cStringIO import StringIO
except ImportError:
from StringIO import StringIO #Py3
import scipy.signal as ss
import ttslab
from ttslab import hrg
from ttslab.trackfile import Track
ttslab.extend(hrg.Utterance, "ufuncs_analysis")
ttslab.extend(Track, "ttslab.trackfile.funcs.tfuncs_praat")
def itemrepr(item):
if "tone" in item:
tone = item["tone"].encode('ascii', 'ignore')
name = unicodedata.normalize("NFKD",
item["name"]).encode('ascii', 'ignore')
if tone is not None and name == "syl":
return tone
else:
return name
else:
return unicodedata.normalize("NFKD",
item["name"]).encode('ascii', 'ignore')
especially if the voice is slightly hoarse...
"""
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import sys
import array
import math
import numpy as np
import ttslab
from ttslab.trackfile import Track
ttslab.extend(Track, "ttslab.trackfile.funcs.tfuncs_praat")
def friendly_log(f):
try:
return math.log(f)
except ValueError:
return float('-1e+10')
if __name__ == "__main__":
fn = sys.argv[1]
outfn = sys.argv[2]
minf0 = float(sys.argv[3])
maxf0 = float(sys.argv[4])
t = Track()
t.get_f0(fn, minpitch=minf0, maxpitch=maxf0, timestep=0.005, fixocterrs=True) #timestep hardcoded here because of hack below...
# -*- coding: utf-8 -*-
""" Implements a number of functions used to analyse utterances
(research purposes)...
"""
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import shutil, os
from tempfile import mkdtemp
from collections import OrderedDict
import ttslab
from ttslab.trackfile import Track
ttslab.extend(Track, "tfuncs_analysis")
import speechlabels as sl
NONE_WORD = "NONE"
WAV_EXT = "wav"
FEAT_EXT = "featvecs"
TEXTGRID_EXT = "TextGrid"
SIG2FV = "sig2fv -coefs melcep -delta melcep -melcep_order 12 -fbank_order 24 -shift 0.005 -factor 5.0 -preemph 0.97 -otype est %(inputfile)s -o %(outputfile)s"
def fill_startendtimes(utt):
""" Use 'end' time feature in segments to fill info for other
items in other common relations..
"""
#segments (are contiguous in time)...
# -*- coding: utf-8 -*-
""" Functions to create HTS labels for synthesis...
See: lab_format.pdf in reference HTS training scripts...
DEMITASSE: I have renamed the "syllable stress" feature to "tone"
"""
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import ttslab
import ttslab.hrg as hrg
ttslab.extend(hrg.Item, "ttslab.ifuncs.synthcontext")
NONE_STRING = "xxx"
def float_to_htk_int(string):
""" Converts a string representing a floating point number to an
integer (time in 100ns units)...
"""
try:
return int(round(float(string) * 10000000))
except:
print(string)
raise
def htk_int_to_float(string):
""" Converts a string representing an integer (time in 100ns units)
# -*- coding: utf-8 -*-
""" Functions to create HTS labels for synthesis...
See: lab_format.pdf in reference HTS training scripts...
"""
from __future__ import unicode_literals, division, print_function #Py2
__author__ = "Daniel van Niekerk"
__email__ = "*****@*****.**"
import ttslab
import ttslab.hrg as hrg
ttslab.extend(hrg.Item, "ttslab.funcs.ifuncs_hts")
NONE_STRING = "xxx"
def float_to_htk_int(string):
""" Converts a string representing a floating point number to an
integer (time in 100ns units)...
"""
try:
return int(round(float(string)*10000000))
except:
print(string)
raise
def htk_int_to_float(string):
""" Converts a string representing an integer (time in 100ns units)
to floating point value (time in seconds)...
"""
return float(string) / 10000000.0