azalea/DeepFormants
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[O] Reformat

Browse Source
This commit is contained in:
Azalea (on HyDEV-Daisy)
2022-04-07 15:53:14 -04:00
parent 689e7d5e83
commit d322444f03
1 changed files with 47 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files
extract_features.py
+8 -22
View File
@@ -7,6 +7,7 @@ import os
import math import math
from inaSpeechSegmenter.features import to_wav from inaSpeechSegmenter.features import to_wav
from numba import float32
from scipy.fftpack.realtransforms import dct from scipy.fftpack.realtransforms import dct
from scipy.signal import lfilter, hamming from scipy.signal import lfilter, hamming
from scipy.fftpack import fft, ifft from scipy.fftpack import fft, ifft
@@ -238,8 +239,10 @@ def specPS(input_wav,pitch):
for k, l in zip(specs[0], specs[1]): for k, l in zip(specs[0], specs[1]):
m = math.sqrt((k ** 2) + (l ** 2)) m = math.sqrt((k ** 2) + (l ** 2))
if m > 0: m = math.log(m) if m > 0: m = math.log(m)
if m == 0: m = epsilon if m == 0:
elif m < 0: m = np.nan m = epsilon
elif m < 0:
m = np.nan
peri.append(m) peri.append(m)
# Filter the spectrum through the triangle filterbank # Filter the spectrum through the triangle filterbank
mspec = np.log10(peri) mspec = np.log10(peri)
@@ -248,12 +251,12 @@ def specPS(input_wav,pitch):
return ceps[:50] return ceps[:50]
def build_single_feature_row(data, Atal): def build_single_feature_row(data: float32[:], Atal):
lpcs = [8, 9, 10, 11, 12, 13, 14, 15, 16, 17] lpc_orders = np.array([8, 9, 10, 11, 12, 13, 14, 15, 16, 17])
arr = [] arr = []
periodo = specPS(data, 50) periodo = specPS(data, 50)
arr.extend(periodo) arr.extend(periodo)
for j in lpcs: for j in lpc_orders:
if Atal: if Atal:
ars = arspecs(data, j, Atal=True) ars = arspecs(data, j, Atal=True)
else: else:
@@ -284,20 +287,3 @@ def create_features(input_wav_filename, feature_filename, begin=None, end=None,
os.remove(wav) os.remove(wav)
return arcep_mat return arcep_mat
if __name__ == "__main__":
# parse arguments
parser = argparse.ArgumentParser(description='Extract features for formants estimation.')
parser.add_argument('wav_file', default='', help="WAV audio filename (single vowel or an whole utternace)")
parser.add_argument('feature_file', default='', help="output feature text file")
parser.add_argument('--begin', help="beginning time in the WAV file", default=0.0, type=float)
parser.add_argument('--end', help="end time in the WAV file", default=-1.0, type=float)
args = parser.parse_args()
if args.begin > 0.0 or args.end > 0.0:
create_features(args.wav_file, args.feature_file, args.begin, args.end)
else:
create_features(args.wav_file, args.feature_file)