def main_deepspeech(args):
args = parse_args_deep() if args is None else args
print('Loading model from file %s' % (args.model), file=sys.stderr)
model_load_start = timer()
ds = Model(args.model, N_FEATURES, N_CONTEXT, args.alphabet, BEAM_WIDTH)
model_load_end = timer() - model_load_start
print('Loaded model in %0.3fs.' % (model_load_end), file=sys.stderr)
if args.lm and args.trie:
print('Loading language model from files %s %s' % (args.lm, args.trie),
file=sys.stderr)
lm_load_start = timer()
ds.enableDecoderWithLM(args.alphabet, args.lm, args.trie, LM_WEIGHT,
WORD_COUNT_WEIGHT, VALID_WORD_COUNT_WEIGHT)
lm_load_end = timer() - lm_load_start
print('Loaded language model in %0.3fs.' % (lm_load_end),
file=sys.stderr)
fs, audio = wave.read(args.audio)
# We can assume 16kHz
audio_length = len(audio) * (1 / 16000)
assert fs == 16000, "Only 16000Hz input WAV files are supported for now!"
print('Running inference.', file=sys.stderr)
inference_start = timer()
print(ds.stt(audio, fs))
inference_end = timer() - inference_start
print('Inference took %0.3fs for %0.3fs audio file.' %
(inference_end, audio_length),
file=sys.stderr)
def predict(self):
try:
#filename如果未定义则会抛出异常
if (filename[-3:] == 'wav'):
fs, wav_data = wave.read(filename)
print(fs)
wav_length = len(wav_data)
print("length of wave:", wav_length)
print("data of wav(int):", type(wav_data))
y = wav_data
segment_sample_num = 4240
segment_move = 400
start = 0
end = start + segment_sample_num
seg_num = 0
while end < wav_length:
start = start + segment_move
end = start + segment_sample_num
seg_num += 1
print(seg_num)
after_segment = [""] * seg_num
start = 0
x_test = np.zeros((seg_num, 32, 129))
for i in range(seg_num):
after_segment[i] = y[start:start + segment_sample_num + 1]
start = start + segment_move
f, t, spectro = sig.spectrogram(after_segment[i], fs,
('hamming'), 256, 128, 256)
#print(f,t)
#print(shape(spectro))
spectro = log(1 + abs(spectro))
for a in range(129):
for b in range(32):
x_test[i][b][a] = spectro[a][b]
print(shape(x_test))
if K.image_data_format() == 'channels_first':
x_test = x_test.reshape(x_test.shape[0], 1, 32, 129)
else:
x_test = x_test.reshape(x_test.shape[0], 32, 129, 1)
x_test = x_test.astype('float32')
print('x_train shape:', x_test.shape)
print(x_test.shape[0], 'test samples')
preds = self.model.predict(x_test, batch_size=32, verbose=1)
print(type(preds))
print(preds)
print(shape(preds))
preds = preds.sum(axis=0)
print(preds)
pred_sum = preds.tolist()
condition = pred_sum.index(max(pred_sum))
if condition == 0:
showinfo('预测', '')
else:
showwarning('warning', '请选择音频文件')
except NameError:
showwarning('warning', '请选择文件')
else:
#print(predictss)
showinfo('The answer of prediction')
import wave, struct
[Fs, x] = wave.read('test.wav', 'r')
print Fs
"""
length = waveFile.getnframes()
for i in range(0,20):
waveData = waveFile.readframes(1)
data = struct.unpack("<h", waveData)
print(int(data[0]))
"""
import wave, struct
w = wave.read('test.wav', 'r')
length = waveFile.getnframes()
for i in range(0, 20):
waveData = waveFile.readframes(1)
data = struct.unpack("<h", waveData)
print(int(data[0]))
from __future__ import print_function
import wave
import struct
from scipy import *
import numpy as np
import scipy.io as scio
import scipy.io.wavfile as wave
from scipy import signal as sig
import os
path = 'H:\RawData2'
mylist = os.listdir(path)
mark=0
for index,y_name in enumerate(mylist):
fs,wav_data = wave.read(path+'\\' +y_name,mmap=False)
print(fs)
wav_length = len(wav_data)
print("length of wave:",wav_length)
print("data of wav(int):",type(wav_data))
y=wav_data
segment_sample_num = 4240
segment_move = 400
start = 0
end = start+segment_sample_num
seg_num = 0
while end < wav_length :
start = start+segment_move
end = start+segment_sample_num
seg_num += 1
def playFile(waveFileName):
if os.path.dirname(waveFileName) == '':
waveFileName = os.path.join(os.path.abspath(os.path.dirname(__file__)),
waveFileName)
with open(waveFileName, 'rb') as wave:
play(wave.read(500 * 1024))