segment.py

   1 import sys
   2 import pympi
   3 import random
   4 import glob
   5 import re
   6 import os
   7
   8 testset = 0.10
   9 data = []
  10
  11
  12 def process(num):
  13     num = re.match('^.*/(\\d+).TextGrid$', num).group(1)
  14     tg = 'textgrid/{:02d}.TextGrid'.format(int(num))
  15
  16     tgob = pympi.TextGrid(tg)
  17     intervalit = tgob.get_tier('lyrics').get_intervals(sort=True)
  18
  19     with open('mfcc/{:02d}.mfcc'.format(int(num))) as mfcc:
  20         (s, e, v) = next(intervalit)
  21         currentframe = 0.0
  22         for l in mfcc:
  23             # Go to next interval
  24             if currentframe > e:
  25                 try:
  26                     (s, e, v) = next(intervalit)
  27                 except StopIteration:
  28                     pass
  29
  30             label = 1 if v == '' else 0
  31             data.append([label] + l.split('\t'))
  32
  33             # Increase time
  34             currentframe += 0.01
  35
  36
  37 if __name__ == '__main__':
  38     datafiles = []
  39     for fl in glob.glob(os.path.join(os.getcwd(), 'textgrid', '*.TextGrid')):
  40         sys.stderr.write('Segment {}\n'.format(fl))
  41         process(fl)
  42
  43     sys.stderr.write('Shuffling {} samples\n'.format(len(data)))
  44     random.shuffle(data)
  45
  46     splitpoint = int(len(data)*testset)
  47     testset = data[:splitpoint]
  48     trainset = data[splitpoint:]
  49     del(data)
  50
  51     sys.stderr.write('Write testset: {} items\n'.format(splitpoint))
  52     with open('test.txt', 'w') as f:
  53         for d in testset:
  54             f.write('\t'.join(map(str, d)))
  55
  56     sys.stderr.write('Write trainingset: {:d} items\n'.format(9*splitpoint))
  57     with open('train.txt', 'w') as f:
  58         for d in trainset:
  59             f.write('\t'.join(map(str, d)))
  60     f.close()