forked from razor1179/pytorch-kaldi-CGS
-
Notifications
You must be signed in to change notification settings - Fork 0
/
save_raw_fea.py
121 lines (88 loc) · 3.65 KB
/
save_raw_fea.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
##########################################################
# pytorch-kaldi v.0.1
# Mirco Ravanelli, Titouan Parcollet
# Mila, University of Montreal
# October 2018
#
# Description: This script generates kaldi ark files containing raw features.
# The file list must be a file containing "snt_id file.wav".
# Note that only wav files are supported here (sphere or other format are not supported)
##########################################################
import scipy.io.wavfile
import math
import numpy as np
import os
from data_io import read_vec_int_ark,write_mat
# Run it for all the data chunks (e.g., train, dev, test) => uncomment
lab_folder='quick_test/dnn4_pretrain-dbn_dnn_ali'
lab_opts='ali-to-pdf'
out_folder='raw_TIMIT_200ms/train'
wav_lst='/home/mirco/pytorch-kaldi-new/quick_test/data/train/wav_lst.scp'
scp_file_out='quick_test/data/train/feats_raw.scp'
#lab_folder='quick_test/dnn4_pretrain-dbn_dnn_ali_dev'
#lab_opts='ali-to-pdf'
#out_folder='raw_TIMIT_200ms/dev'
#wav_lst='/home/mirco/pytorch-kaldi-new/quick_test/data/dev/wav_lst.scp'
#scp_file_out='quick_test/data/dev/feats_raw.scp'
#lab_folder='quick_test/dnn4_pretrain-dbn_dnn_ali_test'
#lab_opts='ali-to-pdf'
#out_folder='raw_TIMIT_200ms/test'
#wav_lst='/home/mirco/pytorch-kaldi-new/quick_test/data/test/wav_lst.scp'
#scp_file_out='quick_test/data/test/feats_raw.scp'
sig_fs=16000 # Hz
sig_wlen=200 # ms
lab_fs=16000 #Hz
lab_wlen=25 #ms
lab_wshift=10 #ms
sig_wlen_samp=int((sig_fs*sig_wlen)/1000)
lab_wlen_samp=int((lab_fs*lab_wlen)/1000)
lab_wshift_samp=int((lab_fs*lab_wshift)/1000)
# Create the output folder
try:
os.stat(out_folder)
except:
os.makedirs(out_folder)
# Creare the scp file
scp_file = open(scp_file_out,"w")
# reading the labels
lab= { k:v for k,v in read_vec_int_ark('gunzip -c '+lab_folder+'/ali*.gz | '+lab_opts+' '+lab_folder+'/final.mdl ark:- ark:-|')}
# reading the list file
with open(wav_lst) as f:
sig_lst = f.readlines()
sig_lst = [x.strip() for x in sig_lst]
for sig_file in sig_lst:
sig_id=sig_file.split(' ')[0]
sig_path=sig_file.split(' ')[1]
[fs,signal]=scipy.io.wavfile.read(sig_path)
signal=signal.astype(float)/32768
signal=signal/np.max(np.abs(signal))
cnt_fr=0
beg_samp=0
frame_all=[]
while beg_samp+lab_wlen_samp<signal.shape[0]:
sample_fr=np.zeros(sig_wlen_samp)
central_sample_lab=int(((beg_samp+lab_wlen_samp/2)-1))
central_fr_index=int(((sig_wlen_samp/2)-1))
beg_signal_fr=int(central_sample_lab-(sig_wlen_samp/2))
end_signal_fr=int(central_sample_lab+(sig_wlen_samp/2))
if beg_signal_fr>=0 and end_signal_fr<=signal.shape[0]:
sample_fr=signal[beg_signal_fr:end_signal_fr]
else:
if beg_signal_fr<0:
n_left_samples=central_sample_lab
sample_fr[central_fr_index-n_left_samples+1:]=signal[0:end_signal_fr]
if end_signal_fr>signal.shape[0]:
n_right_samples=signal.shape[0]-central_sample_lab
sample_fr[0:central_fr_index+n_right_samples+1]=signal[beg_signal_fr:]
frame_all.append(sample_fr)
cnt_fr=cnt_fr+1
beg_samp=beg_samp+lab_wshift_samp
frame_all=np.asarray(frame_all)
# Save the matrix into a kaldi ark
out_file=out_folder+'/'+sig_id+'.ark'
write_mat(out_file, frame_all, key=sig_id)
print(sig_id)
scp_file.write(sig_id+' '+out_folder+'/'+sig_id+'.ark:'+str(len(sig_id)+1)+'\n')
N_fr_comp=1 + math.floor((signal.shape[0] - 400) / 160)
#print("%s %i %i "%(lab[sig_id].shape[0],N_fr_comp,cnt_fr))
scp_file.close()