def gerador_de_caracteriticas():
all_features = []
for base in bases_B:
features_list = []
files.append(func.get_filenames(path=base, filetype='.wav'))
for i in range(len(files[-1])):
instance, rate = librosa.load(base + '/' + files[-1][i])
features = func.extract_feature(instance, rate)
# rotulacao
if base == btraning_normal:
features.append("Normal")
elif base == btraning_mumur:
features.append("Murmur")
elif base == btraining_extrastole:
features.append("Extrastole")
features_list.append(features)
all_features.append(features)
print('Done with: ' + files[-1][i] + " - " + features[6])
# escrita em arquivo CSV
dir_db = os.path.basename(base)
func.write_csv('librosa_features_of_' + dir_db, features_list)
print('Done with: ' + base + '\n')
os.chdir(path_B)
func.write_csv("BaseB_librosa_features", all_features)
def main_a():
files = []
all_features = []
for base in bases_A:
instances_list = []
features_list = []
filtered_instances_list = []
files.append(func.get_filenames(
path=base, filetype='.wav')) # gera array com nome dos audios
for i in range(len(files[-1])):
# instance, rate = func.load_sound_files(base + '/' + files[-1][i])
# rate, instance = wavfile.read(base + '/' + files[-1][i])
instance, rate = librosa.load(base + '/' + files[-1][i])
# [rate, instance] = audioBasicIO.readAudioFile(base + '/' + files[-1][i])
instances_list.append(instance)
# filtragem e decomposicao
rec_signal = func.wavelet_filtering(instance, th, rate)
filtered_instances_list.append(rec_signal)
# extracao de caracteriticas
features = func.extract_feature(instance, rate) # librosa
# features = audioFeatureExtraction.stFeatureExtraction(instance, rate, 0.5 * rate, 0.25 * rate)
# features = audioFeatureExtraction.mtFeatureExtraction(
# instance, rate, 0.5 * rate, 0.5 * rate,0.25 * rate, 0.25 * rate
# )
# rotulacao
if base == atraning_normal:
features.append("Normal")
elif base == atraning_mumur:
features.append("Murmur")
elif base == atraining_extrahls:
features.append("Extra Heart Sound")
elif base == atraining_artifact:
features.append("Artifact")
features_list.append(features)
all_features.append(features)
print('Done with: ' + files[-1][i])
# escritas no CVS
dir_db = os.path.basename(base)
func.write_csv('librosa_filtered_features_of_' + dir_db, features_list)
print('Done with: ' + base + '\n')
os.chdir(path_A)
func.write_csv("BaseA_librosa_filtered_features", all_features)
print 'Done!'
def test_write_csv():
"""
Test procedure for the function write_csv()
"""
print('Testing write_csv()')
# Access the file relative to this one, not the user's terminal
parent = os.path.split(__file__)[0]
# First test (erase any existing file)
fpath = os.path.join(parent, 'files', 'temp1.csv')
open(fpath, 'w').close()
func.write_csv(FILE1, fpath)
# Check file was created
introcs.assert_true(os.path.exists(fpath))
file = open(fpath)
actual = file.read()
file.close()
file = open(os.path.join(parent, 'files', 'writecsv1.csv'))
correct = file.read()
file.close()
# Check to see if they are the same
introcs.assert_equals(correct, actual)
# Second test (erase any existing file)
fpath = os.path.join(parent, 'files', 'temp2.csv')
open(fpath, 'w').close()
func.write_csv(FILE2, fpath)
# Check file was created
introcs.assert_true(os.path.exists(fpath))
file = open(fpath)
actual = file.read()
file.close()
file = open(os.path.join(parent, 'files', 'writecsv2.csv'))
correct = file.read()
file.close()
# Check to see if they are the same
introcs.assert_equals(correct, actual)
def main_pn():
files = []
labels = []
for base in bases_Physioef:
instances_list = []
features_list = []
files.append(func.get_filenames(path=base, filetype='.wav'))
labels.append(func.get_filenames(path=base, filetype='.hea'))
for i in range(len(files[-1])):
# instance, rate = func.load_sound_files(base + '/' + files[-1][i])
# rate, instance = wavfile.read(base + '/' + files[-1][i])
instance, rate = librosa.load(base + '/' + files[-1][i])
# filtragem e decomposicao
rec_signal = func.wavelet_filtering(instance, th)
instances_list.append(rec_signal)
# extracao de caracteriticas
features = func.extract_feature(rec_signal, rate)
# rotulacao physionet
label_object = open(base + '/' + labels[-1][i], 'r')
label = label_object.read().split('\n')[-2]
features.append(label)
# features = func.pyAudioAnalysis_features(recSignal, rate)
features_list.append(features)
print('Done with: ' + files[-1][i])
func.write_csv(path_Physio + base[-1] + '_filtered', features_list)
# dir_db = os.path.basename(base)
# func.write_csv('librosa_features_of_' + dir_db, features_list)
print('Done with: ' + base + '\n')
print 'Done!'
def main():
files = []
# instance = []
labels = []
file_paths = []
files = func.get_filenames(mumur, '.wav')
# carrega arquivos de audio
for s in files:
file_paths.append(mumur + '/' + s)
instance, rates = func.load_sound_files(file_paths)
# ecg = pywt.data.ecg() # dataset do pywavelets
# Wavelet Daubechies order 6 (ortogonal)
wavelet = pywt.Wavelet('db6')
(phi, psi, x) = wavelet.wavefun()
# wavelet = signal.daub(6)
# identification
print "Rotulo:", label[1]
print "Nome Arquivo:", files[0]
print "Frequencia (fs): ", rates[0]
# print "\n", wavelet
# filtragem e decomposicao
decimateSignal = signal.decimate(instance[0], 10) # decomposicao de sinal
# y = func.butter_lowpass_filter(decimateSignal, lowcut, fs, order=6) # passa baixa
coeffs = pywt.wavedec(decimateSignal, 'db6',
level=4) # transformada wavelet
cA4, cD4, cD3, cD2, cD1 = coeffs
cD4 = pywt.threshold(cD4, th, mode='less')
cD3 = pywt.threshold(cD3, th, mode='less')
cD2 = pywt.threshold(cD2, th, mode='less')
cD1 = pywt.threshold(cD1, th, mode='less')
coeffs = cA4, cD4, cD3, cD2, cD1
# new_coeffs = []
# aux = []
# for cd in coeffs[1:]:
# for d in cd:
# if 288.0 >= d > 0.0:
# aux = np.append(aux, d)
# new_coeffs.append(aux)
# reconstrucao do sinal
recSignal = pywt.waverec(coeffs, 'db6')
# verifica se sinais sao iguais
if np.array_equal(recSignal, instance[0]):
print('! sinais iguais !')
# smooth = signal.savgol_filter(recSignal, 5, 2)
# print smooth
# z = np.polyfit(smooth, , 3)
# print z
# std = np.std(smooth)
# print "std", std
# helper = []
# for i, x in enumerate(smooth):
# if x > std:
# helper.append(i)
#
# print helper
# Spectrogram
# f, Pwelch_spec = signal.welch(recSignal, fs, scaling='spectrum')
# plt.semilogy(f, Pwelch_spec)
# plt.xlabel('frequency [Hz]')
# plt.ylabel('PSD')
# plt.grid()
# plt.show()
#
# f = pywt.threshold(f, 195, mode='less')
# plt.semilogy(f, Pwelch_spec)
# plt.show()
#
# spectro = aF.stSpectogram(recSignal, rates[0], 0.05*rates[0], 0.025*rates[0], PLOT=True)
# print spectro
func.write_csv(files[0], recSignal)
# Plotagem dos sinais
func.plot_imagens(psi, 'Wavelet Daubechies de Ordem 6') # wavelet
func.plot_imagens(instance[0], 'sinal nao filtrado')
# func.plot_imagens(decimateSignal, 'sinal reduzido - fator 10')
func.plot_imagens(recSignal, 'sinal reconstruido')
# func.plot_imagens(recSignal2, 'sinal reconstruido - upcoef')
# func.plot_imagens(smooth, 'suavizacao triangular')
# func.plot_imagens(z,'polyfit')
plt.show()