import feature_extractor as fe
from sys import argv
import os
"""
Put all samples (mfcc, deltas and delta deltas) of speaker n to file samples_n.txt. Repeat for all speakers.
By uncommenting the lines with i, number skipping will be avoided in resulting files,
but the file names will not match the correct speaker.
"""
if __name__ == '__main__':
i = 1
for x in range(155):
print "Speaker {} in progress.".format(x + 225)
pathname = str(x + 225)
#outfile = "samples_{}".format(pathname)
outfile = "samples_{}".format(i)
if os.path.exists("VCTK-Corpus\wav48\p"+pathname):
for y in range(510):
if y < 10:
z = "00"
elif y < 100:
z = "0"
else:
z = ""
name = pathname + "_" + str(y)
filename = "VCTK-Corpus\wav48\p"+pathname+"\p"+pathname+"_"+z+str(y)+".wav"
if os.path.exists(filename):
fe.extract_and_save(filename, name, outfile, "mfcc", 2)
i += 1
ystart = 50
yend = 52
for x in range(xstart,xend):
pathname = str(x + 225)
if (outfile == "train" and x > ubm_x):
name = ".ubm"
else:
name = pathname
for y in range(ystart, yend):
if y < 10:
z = "00"
elif y < 100:
z = "0"
else:
z = ""
if (outfile == "test"):
name = pathname + "_" + str(y)
filename = "VCTK-Corpus\wav48\p"+pathname+"\p"+pathname+"_"+z+str(y)+".wav"
if os.path.exists(filename):
if (remove_silence != 0):
sr.remove_silence(filename)
filename = filename + ".sr"
fe.extract_and_save(filename, name, outfile, type, delta_amount)
print filename
os.remove(filename)
else:
print filename
fe.extract_and_save(filename, name, outfile, type, delta_amount)
