def GetData(self, n_start, n_amount=32, dataType='same'):
assert (n_amount % CLASS_NUM == 0)
if dataType == self.pathSameLabel:
path = self.listSame[n_start][0]
data_label = np.array([self.listSame[n_start][1]])
elif dataType == self.pathDistinctLabel:
path = self.listDistinct[n_start][0]
data_label = np.array([self.listDistinct[n_start][1]])
wavsignal, fs = read_wav_data(path)
if self.feature_type in [
'spec', 'Spec', 'SPEC', 'Spectrogram', 'SPECTROGRAM'
]:
data_input = GetFrequencyFeatures(wavsignal,
fs,
self.feat_dimension,
self.frame_length,
shift=160)
elif self.feature_type in ['mfcc', 'MFCC', 'Mfcc']:
data_input = SimpleMfccFeatures(wavsignal, fs)
else:
print('Unknown feature type.')
assert (0)
data_input = data_input.reshape(data_input.shape[0],
data_input.shape[1], 1)
return data_input, data_label
def skstyleDataset(input_folder, check_num, stft_spec=False):
assert (isinstance(check_num, int))
feature = []
target = []
folders = os.listdir(input_folder)
# place a self-check
if len(folders) != check_num:
print('error occurred when cheking subject numbers.')
assert (0)
else:
for folder in folders:
subf_bowel = os.path.join(input_folder, folder, 'bowels')
files = os.listdir(subf_bowel)
if len(files) in (200, 198):
for file in files:
path = os.path.join(subf_bowel, file)
wavsignal, fs = read_wav_data(path)
if stft_spec == True:
data_input = GetFrequencyFeatures(wavsignal,
fs,
200,
400,
shift=160)
else:
data_input = SimpleMfccFeatures(wavsignal, fs)
data_input = np.reshape(data_input, [-1])
feature.append(data_input)
target.append(1)
subf_non = os.path.join(input_folder, folder, 'non')
files = os.listdir(subf_non)
if len(files) in (200, 198):
for file in files:
path = os.path.join(subf_non, file)
wavsignal, fs = read_wav_data(path)
if stft_spec == True:
data_input = GetFrequencyFeatures(wavsignal,
fs,
200,
400,
shift=160)
else:
data_input = SimpleMfccFeatures(wavsignal, fs)
data_input = np.reshape(data_input, [-1])
feature.append(data_input)
target.append(0)
else:
print('wrong wavfile number.')
assert (0)
return np.array(feature), np.array(target)
set_session(tf.Session(config=config))
#1. load image
three = []
results =[]
spath = os.path.join(os.getcwd(),'log&&materials','105660089020341_2018_11_13_08_20_07.wav')
three.append(spath)
spath = os.path.join(os.getcwd(),'log&&materials','20843728498872_2018_01_08_10_42_58.wav')
three.append(spath)
spath = os.path.join(os.getcwd(),'log&&materials','199481741554988_2018_10_20_00_04_25.wav')
three.append(spath)
# n1path = os.path.join(os.getcwd(),'log&&materials','noisy_sounds','199481741554988_2018_10_20_00_00_00.wav')
for spath in three:
signal,fs = read_wav_data(spath)
spec = GetFrequencyFeatures(signal,fs)
img = spec.reshape(1,spec.shape[0], spec.shape[1], 1)
#2. load model
# mpath = os.path.join(os.getcwd(),'CNNevaluation','residual','resi=1','SPEC_Residual_8_epoch14.h5')
# mpath = os.path.join(os.getcwd(),'CNNdesign','spec+inception','10','SPEC_Inception_10_epoch4.h5')
mpath = os.path.join(os.getcwd(),'CNNevaluation','regular','regi=0','spec_Regular_8_epoch12.h5')
model = load_model(mpath,custom_objects={'FocalLoss': focal_loss,'focal_loss_fixed': focal_loss()})
optimizer = optimizers.Adadelta()
model.compile(optimizer=optimizer, loss=[focal_loss(alpha=0.25, gamma=2)])
#3. try to predictverbose=0
print(np.argmax(model.predict(img)))
print(model.predict(img,verbose=1))
# print(np.argmax(model.predict(img)[0]))
# print(model.layers[27])
# print(model.layers[28])
# print(model.layers[29])
def GetData(self, n_start, n_amount=32, mode='balanced'):
'''
读取数据，返回神经网络输入值和输出值矩阵(可直接用于神经网络训练的那种)
参数：
n_start：从编号为n_start数据开始选取数据
n_amount：选取的数据数量，默认为1，即一次一个wav文件
返回：
四个音频四个label，
'''
# 随机从四个文件夹中拿一条数据，判断是否大于1s，否就重拿
assert (n_amount % CLASS_NUM == 0)
category = (self.list_bowel1, self.list_non0)
link = ('bowels', 'non')
label = (1, 0)
if self.feature_type in [
'spec', 'Spec', 'SPEC', 'Spectrogram', 'SPECTROGRAM'
]:
if mode == 'balanced':
data = []
labels = []
for genre in range(CLASS_NUM):
for file in range(n_amount // CLASS_NUM):
filename = category[genre][(n_start + file) %
self.DataNum[genre]]
# filename = category[genre][(n_start + file) % min(self.DataNum)]
path = self.common_path + link[
genre] + self.slash + filename
wavsignal, fs = read_wav_data(path)
# data_input = SimpleMfccFeatures(wavsignal, fs)
data_input = GetFrequencyFeatures(wavsignal,
fs,
self.feat_dimension,
self.frame_length,
shift=160)
data_input = self.shifting(data_input)
data_input = data_input.reshape(
data_input.shape[0], data_input.shape[1], 1)
data.append(data_input)
data_label = np.array([label[genre]])
labels.append(data_label)
return data, labels
if mode == 'non-repetitive':
path = self.list_path[n_start][0]
data_label = np.array([self.list_path[n_start][1]])
wavsignal, fs = read_wav_data(path)
data_input = GetFrequencyFeatures(wavsignal,
fs,
self.feat_dimension,
self.frame_length,
shift=160)
data_input = data_input.reshape(data_input.shape[0],
data_input.shape[1], 1)
return data_input, data_label
elif self.feature_type in ['mfcc', 'MFCC', 'Mfcc']:
if mode == 'balanced':
data = []
labels = []
for genre in range(CLASS_NUM):
for file in range(n_amount // CLASS_NUM):
filename = category[genre][(n_start + file) %
self.DataNum[genre]]
# filename = category[genre][(n_start + file) % min(self.DataNum)]
path = self.common_path + link[
genre] + self.slash + filename
wavsignal, fs = read_wav_data(path)
data_input = SimpleMfccFeatures(wavsignal, fs)
data_label = np.array([label[genre]])
data_input = self.shifting(data_input)
data_input = data_input.reshape(
data_input.shape[0], data_input.shape[1], 1)
data.append(data_input)
labels.append(data_label)
return data, labels
if mode == 'non-repetitive':
path = self.list_path[n_start][0]
data_label = np.array([self.list_path[n_start][1]])
wavsignal, fs = read_wav_data(path)
data_input = SimpleMfccFeatures(wavsignal, fs)
data_input = data_input.reshape(data_input.shape[0],
data_input.shape[1], 1)
return data_input, data_label
else:
print('Unknown feature type.')
assert (0)