-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmelSpectrogram.py
142 lines (132 loc) · 6.22 KB
/
melSpectrogram.py
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
138
139
140
141
142
# code written by Noh Hyun-kyu POSTECH Oct 2019
# -*- coding: utf-8 -*-
import random
import librosa
import librosa.display
import numpy as np
import matplotlib.pyplot as plt
import scipy.signal
import os
import normalize_in_time
import struct
frame_length = 0.020
frame_stride = 0.010
testset_ratio = 0.2
foldername = '../dataset_students'
def generate_spectrogram(wav_file):
y, sr = librosa.load(wav_file, sr=48000)
#print(keyword,': file_name=',file_name,end="\t") # HJP
y = normalize_in_time.normalize_array(y)
input_nfft = int(round(sr*frame_length))
input_stride = int(round(sr*frame_stride))
S = librosa.feature.melspectrogram(y=y, n_mels=40, n_fft=input_nfft, hop_length=input_stride, window=scipy.signal.windows.hann)
S_log = np.log10(S + 1e-5)
S_log_100 = normalize_in_time.normalize_in_time(S_log,y,S,2)
return S_log_100
def WriteTrainingSet(Small_mode = False):
keywords = ['ALEXA', 'BIXBY', 'GOOGLE', 'JINIYA', 'KLOVA']
X_training = []
Y_training = []
X_test = []
Y_test = []
labels = 0
os.system('del ' + foldername + '/' + 'train_data.bin')
os.system('del ' + foldername + '/' + 'train_label.bin')
os.system('del ' + foldername + '/' + 'test_data.bin')
os.system('del ' + foldername + '/' + 'test_label.bin')
fp_train_data = open(foldername + '/' + 'train_data.bin','wb')
fp_train_label = open(foldername + '/' + 'train_label.bin','wb')
fp_test_data = open(foldername + '/' + 'test_data.bin','wb')
fp_test_label = open(foldername + '/' + 'test_label.bin','wb')
if Small_mode :
for keyword in keywords:
wav_folder = foldername + '/' + keyword
file_observed = os.listdir(wav_folder)
random.shuffle(file_observed)
print('converting ',file_observed,' files of ', keyword, ':')
test_num = int(len(file_observed) * testset_ratio)
fp_train_data.write(struct.pack('i', len(file_observed) - test_num))
fp_train_label.write(struct.pack('i' ,len(file_observed) - test_num))
fp_test_data.write(struct.pack('i', test_num))
fp_test_label.write(struct.pack('i' ,test_num))
idx = 0
for file_name in file_observed:
wav_file = wav_folder + '/' + file_name
S_log_100 = generate_spectrogram(wav_file)
if(idx < test_num) :
fp_test_data.write(struct.pack('d'*100 * 40, *S_log_100.flatten() ))
fp_test_label.write(struct.pack('i' , labels))
else :
fp_train_data.write(struct.pack('d'*100 * 40, *S_log_100.flatten() ))
fp_train_label.write(struct.pack('i' , labels))
idx = idx + 1
labels = labels + 1
else :
for keyword in keywords:
wav_folder = foldername + '/train/' + keyword
file_observed = os.listdir(wav_folder)
file_observed.sort()
print('converting ',file_observed,' files of ', keyword, ':')
fp_train_data.write(struct.pack('i', len(file_observed)))
fp_train_label.write(struct.pack('i' ,len(file_observed)))
for file_name in file_observed:
wav_file = wav_folder + '/' + file_name
S_log_100 = generate_spectrogram(wav_file)
fp_train_data.write(struct.pack('d'*100 * 40, *S_log_100.flatten() ))
fp_train_label.write(struct.pack('i' , labels))
title_str = keyword + file_name
plt.title(title_str)
labels = labels + 1
labels = 0
for keyword in keywords:
wav_folder = foldername + '/test/' + keyword
file_observed = os.listdir(wav_folder)
file_observed.sort()
print('converting ',file_observed,' files of ', keyword, ':')
fp_test_data.write(struct.pack('i', len(file_observed)))
fp_test_label.write(struct.pack('i' ,len(file_observed)))
for file_name in file_observed:
wav_file = wav_folder + '/' + file_name
S_log_100 = generate_spectrogram(wav_file)
fp_test_data.write(struct.pack('d'*100*40 , *S_log_100.flatten() ))
fp_test_label.write(struct.pack('i' , labels))
title_str = keyword + file_name
plt.title(title_str)
labels = labels + 1
fp_train_label.close()
fp_test_label.close()
fp_train_data.close()
fp_test_data.close()
def ReadTrainingSet():
fp_train_data = open(foldername + '/' + 'train_data.bin','rb')
fp_train_label = open(foldername + '/' + 'train_label.bin','rb')
fp_test_data = open(foldername + '/' + 'test_data.bin','rb')
fp_test_label = open(foldername + '/' + 'test_label.bin','rb')
X_training = []
Y_training = []
X_test = []
Y_test = []
keywords = ['ALEXA', 'BIXBY', 'GOOGLE', 'JINIYA', 'KLOVA']
for keyword in keywords:
file_observed = (struct.unpack('i', fp_train_data.read(4)))[0]
file_observed = (struct.unpack('i', fp_train_label.read(4)))[0]
print('Train : ' + str(file_observed))
for i in range(file_observed):
S_log_100 = np.array(struct.unpack('d'* 100 * 40,fp_train_data.read(8*100*40))).reshape((40, 100))
labels = (struct.unpack('i',fp_train_label.read(4)))[0]
X_training.append(S_log_100)
Y_training.append(labels)
for keyword in keywords:
file_observed = (struct.unpack('i', fp_test_data.read(4)))[0]
file_observed = (struct.unpack('i', fp_test_label.read(4)))[0]
print('Test : ' + str(file_observed))
for i in range(file_observed):
S_log_100 = np.array(struct.unpack('d'* 100 * 40,fp_test_data.read(8*100*40))).reshape((40, 100))
labels = (struct.unpack('i',fp_test_label.read(4)))[0]
X_test.append(S_log_100)
Y_test.append(labels)
fp_train_label.close()
fp_test_label.close()
fp_train_data.close()
fp_test_data.close()
return X_training, Y_training, X_test, Y_test