Word-Boundary-Detection/create_data.py at master · macandro96/Word-Boundary-Detection · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
import numpy as np
import os
import librosa
import soundfile as sf
from pathlib import Path
import pickle
import random

def getGloVector(glove_vectors_file = "glove.6B.50d.txt"):
	glove_wordmap = {}
	with open(glove_vectors_file, "r", encoding="utf8") as glove:
	    for line in glove:
	        name, vector = tuple(line.split(" ", 1))
	        glove_wordmap[name] = np.fromstring(vector, sep=" ")
	return glove_wordmap


def fill_unk(unk):
	global glove_wordmap
	glove_wordmap[unk] = RS.multivariate_normal(m,np.diag(v))
	return glove_wordmap[unk]


glove_wordmap = getGloVector()
wvecs = []
for item in glove_wordmap.items():
    wvecs.append(item[1])

s = np.vstack(wvecs)

# Gather the distribution hyperparameters
v = np.var(s,0)
m = np.mean(s,0)
RS = np.random.RandomState()

def getWordBoundariesAndWords(file_name):
  f = open(file_name, "r")
  content = f.readlines()
  words = list()
  start = list()
  end = list()
  for line in content:
    tokens = line.split()
    start.append(tokens[0])
    end.append(tokens[1])
    words.append(tokens[2].lower())
    # if vocab_dict.get(tokens[2].lower()) is None:
    #   vocab_dict[tokens[2].lower()] = True

  return words, start, end

def getWordVector(word):
	# if word_map.get(word) is None:
	if glove_wordmap.get(word) is None:
		wrd_map = fill_unk(word)
		# word_map[word] = wrd_map
	else:
		wrd_map = glove_wordmap[word]
	return wrd_map


def read_wav(file_name):
  data, samplerate = sf.read(file_name)
  return data

dir_map = {}


def createDirMap():
	count = 0
	X_audio = list()
	X_word = list()
	Y1 = list()
	Y2 = list()
	dir_list = [x[0] for x in os.walk("./test")]
	print(len(dir_list))

	for dire in dir_list:
		folder_lis = [x[0] for x in os.walk(dire+"/")]
		# print(folder_lis)
		for folder in folder_lis:
			file_dict = {}
			# print(dire, folder)
			for filename in os.listdir(folder+"/"):
				if filename.endswith(".wav"):
					key = filename.split(".")
					file_dict[key[0]] = True

			for file_name in file_dict.keys():
				file_dir = folder+"/"+file_name
				wav_file = file_dir+".wav"
				wrd_file = file_dir+".wrd"
				my_file = Path(wrd_file)
				if not my_file.is_file():
					continue
				data = read_wav(wav_file)
				data = np.array([ data[i] for i in sorted(random.sample(range(len(data)), 10000)) ])
				words, start, end = getWordBoundariesAndWords(wrd_file)
				word_lis = list()
				for word in words:
					wrd_vec = getWordVector(word)
					word_lis.append(np.array(wrd_vec))
				#while len(word_lis) < 30:
				#	word_lis.append(np.zeros(50))
				#	start.append(0)
				#	end.append(0)

				X_audio.append(data)

				X_word.append(np.array(word_lis, dtype=np.float32))
				Y1.append(np.array(start, dtype=np.float32))
				Y2.append(np.array(end, dtype=np.float32))
				# print (count)
				# count += 1
	X_audio = np.array(X_audio)
	X_word = np.array(X_word)
	Y1 = np.array(Y1)
	Y2 = np.array(Y2)
	print(X_audio.shape, X_word.shape, Y1.shape, Y2.shape)
	return X_audio, X_word, Y1, Y2


X_audio, X_word, Y1, Y2 = createDirMap()

with open("audio_test.txt","wb") as file:
	pickle.dump(X_audio,file)
with open("word_original_test.txt","wb") as file:
	pickle.dump(X_word,file)

with open("Y1_original_test.txt","wb") as file:
	pickle.dump(Y1,file)

with open("Y2_original_test.txt","wb") as file:
	pickle.dump(Y2,file)