def saveSignalFeatures():
"""
This function saves all the features for each signal into a giant dataframe
This is so we don't have to re-derive the peaks, intervals, etc. for each signal
Parameters
----------
None
Returns
-------
Saves dataframe as hardcoded_features.csv where each row is a filtered signal with the getFeatures() features
"""
records = wave.getRecords('All')[0]
returnMatrix = []
for i in records:
sig = Signal(i, wave.load(i))
features = getFeatures(sig)
returnMatrix.append(features)
df = pd.DataFrame(returnMatrix, )
df.to_csv('hardcoded_features.csv')
def getFeaturesNames():
records = wave.getRecords('All')[0]
returnMatrix = []
sig = Signal(records[0], wave.load(records[0]))
features_names = getFeatures(sig, names=True)
return features_names
def feature_extract():
"""
this function creates a feature matrix from partitioned data
Parameters
----------
None
Returns
-------
A pickle dump of the following:
tuple of tuples:
test (1/10th of data) tuple:
testing subset feature matrix, 2D array
list of record labels N O A ~
list of record names
training (9/10th of data) tuple:
training subset feature matrix, 2D array
list of record labels N O A ~
"""
records_labels = wave.getRecords('All')
partitioned = wave.getPartitionedRecords(0) # partition nth 10th, 0-9
testing = partitioned[0]
training = partitioned[1]
testMatrix = np.array([getFeaturesHardcoded("A00001")])
trainMatrix = np.array([getFeaturesHardcoded("A00001")])
for i in records_labels[0]:
if i in testing[0]:
testMatrix = np.concatenate(
(testMatrix, [getFeaturesHardcoded(i)]))
elif i in training[0]:
trainMatrix = np.concatenate(
(trainMatrix, [getFeaturesHardcoded(i)]))
testMatrix = np.delete(testMatrix, (0),
axis=0) # get rid of A00001 initialization array
trainMatrix = np.delete(trainMatrix, (0), axis=0)
featureMatrix = ((testMatrix, testing[1], testing[0]), (trainMatrix,
training[1]))
pickle.dump(featureMatrix, open("feature_matrices", 'wb'))
records_train = pd.read_csv(path + 'REFERENCE.csv', names=['file', 'answer'])
records_validation = pd.read_csv('validation/' + 'REFERENCE.csv',
names=['file', 'answer'])
new_validation = []
for index, row in records_validation.iterrows():
record = records_train.ix[records_train['file'] == row['file']]
filename = record['file'].tolist()
answer = record['answer'].tolist()
new_validation.append([filename[0], answer[0]])
#print(len(records))
#print(records)
records = wave.getRecords('All')
wired_list = []
feat_list = []
for record in records:
# try:
data = wave.load(record)
print('running record: ' + record)
#sig = challenge.Signal(record, data)
noise_features = challenge.noise_feature_extract(data)
feat_list.append(noise_features)
print('the number of records in the feature list: ' + str(len(feat_list)))
# except:
# wired_list.append(record)
# print ('stupid one found: ' + record)
def pointDetection():
records = wave.getRecords('N') # N O A ~
data = wave.load(records[0])
sig = Signal(records[0], data)
fig = plt.figure(figsize=(200, 6)) # I used figures to customize size
ax = fig.add_subplot(211)
ax.plot(sig.data)
ax.plot(*zip(*sig.Ppeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.Tpeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.RPeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.QPoints), marker='o', color='r', ls='')
ax.plot(*zip(*sig.SPoints), marker='o', color='r', ls='')
ax.axhline(sig.baseline)
ax.set_title(sig.name)
fig.savefig('/Users/samy/Downloads/{0}.png'.format(sig.name))
plt.show()
records = wave.getRecords('A') # N O A ~
data = wave.load(records[0])
sig = Signal(records[0], data)
fig = plt.figure(figsize=(200, 6)) # I used figures to customize size
ax = fig.add_subplot(211)
ax.plot(sig.data)
ax.plot(*zip(*sig.Ppeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.Tpeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.RPeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.QPoints), marker='o', color='r', ls='')
ax.plot(*zip(*sig.SPoints), marker='o', color='r', ls='')
ax.axhline(sig.baseline)
ax.set_title(sig.name)
fig.savefig('/Users/samy/Downloads/{0}.png'.format(sig.name))
plt.show()
records = wave.getRecords('O') # N O A ~
data = wave.load(records[0])
sig = Signal(records[0], data)
fig = plt.figure(figsize=(200, 6)) # I used figures to customize size
ax = fig.add_subplot(211)
ax.plot(sig.data)
ax.plot(*zip(*sig.Ppeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.Tpeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.RPeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.QPoints), marker='o', color='r', ls='')
ax.plot(*zip(*sig.SPoints), marker='o', color='r', ls='')
ax.axhline(sig.baseline)
ax.set_title(sig.name)
fig.savefig('/Users/samy/Downloads/{0}.png'.format(sig.name))
plt.show()
records = wave.getRecords('~') # N O A ~
data = wave.load(records[0])
sig = Signal(records[0], data)
fig = plt.figure(figsize=(200, 6)) # I used figures to customize size
ax = fig.add_subplot(211)
ax.plot(sig.data)
ax.plot(*zip(*sig.Ppeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.Tpeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.RPeaks), marker='o', color='r', ls='')
ax.plot(*zip(*sig.QPoints), marker='o', color='r', ls='')
ax.plot(*zip(*sig.SPoints), marker='o', color='r', ls='')
ax.axhline(sig.baseline)
ax.set_title(sig.name)
fig.savefig('/Users/samy/Downloads/{0}.png'.format(sig.name))
plt.show()