def features_one_file(f):
file_name = os.path.basename(f).split(".")[0]
treatment, animal = file_name.split("_")
pol = polygram_from_pkl(f)
#eegs = decompose_signal(pol["EEG_parietal_cereb"], levels=[1,2,3,4,5,6])
eegs = decompose_signal(pol["EEG_parietal_frontal"], levels=[1,2,3,4,5,6])
emgs = decompose_signal(pol["EMG_REF"],[1,2,3, 4], keep_a=False)
#
pol2 = eegs.merge(pol["EEG_parietal_frontal"])
pol2 = pol2.merge(emgs)
pol2 = pol2.merge(pol["EMG_REF"])
pol2 = pol2.merge(pol["vigilance_state"])
##normalise
pol2 = pol2.map_signal_channels(lambda c : (c - np.mean(c))/ np.std(c))
feature_factory = [
PowerFeatures(),
HjorthFeatures(),
# NonLinearFeatures(),
#
# # FIXME skip for now -> speed
EntropyFeatures(),
FractalFeatures(),
VigilState(),]
all_rows = []
print "processing " + f
old_p = 0
for t, w in pol2.iter_window(WINDOW_SIZE, WINDOW_LAG):
dfs = []
for c in w.channels:
for ff in feature_factory:
feature_vec = ff.make_vector(c)
if not feature_vec is None:
dfs.append(feature_vec)
p = int(100 * t/ pol2.duration.total_seconds())
if p != old_p:
print f, p, "%"
old_p = p
row = pd.concat(dfs, axis=1)
row.index = [t]
all_rows.append(row)
tmp_df = pd.concat(all_rows)
tmp_df["animal"] = animal
tmp_df["treatment"] = treatment
return tmp_df
def features_one_file(f):
file_name = os.path.basename(f).split(".")[0]
treatment, animal = file_name.split("_")
pol = polygram_from_pkl(f)
#eegs = decompose_signal(pol["EEG_parietal_cereb"], levels=[1,2,3,4,5,6])
eegs = decompose_signal(pol["EEG_parietal_frontal"],
levels=[1, 2, 3, 4, 5, 6])
emgs = decompose_signal(pol["EMG_REF"], [1, 2, 3, 4], keep_a=False)
#
pol2 = eegs.merge(pol["EEG_parietal_frontal"])
pol2 = pol2.merge(emgs)
pol2 = pol2.merge(pol["EMG_REF"])
pol2 = pol2.merge(pol["vigilance_state"])
##normalise
pol2 = pol2.map_signal_channels(lambda c: (c - np.mean(c)) / np.std(c))
feature_factory = [
PowerFeatures(),
HjorthFeatures(),
# NonLinearFeatures(),
#
# # FIXME skip for now -> speed
EntropyFeatures(),
FractalFeatures(),
VigilState(),
]
all_rows = []
print "processing " + f
old_p = 0
for t, w in pol2.iter_window(WINDOW_SIZE, WINDOW_LAG):
dfs = []
for c in w.channels:
for ff in feature_factory:
feature_vec = ff.make_vector(c)
if not feature_vec is None:
dfs.append(feature_vec)
p = int(100 * t / pol2.duration.total_seconds())
if p != old_p:
print f, p, "%"
old_p = p
row = pd.concat(dfs, axis=1)
row.index = [t]
all_rows.append(row)
tmp_df = pd.concat(all_rows)
tmp_df["animal"] = animal
tmp_df["treatment"] = treatment
return tmp_df
def data_for_one_file(file, channel_name, dfs):
pol = polygram_from_pkl(file)
for t, w in pol.iter_window(WINDOW_SIZE, WINDOW_LAG):
eeg = w[channel_name]
ann = w["vigilance_state"]
if ann.probas.all() > 0:
y = ann.values[0]
periodo = make_periodogram(eeg)
try:
dfs[y].append(periodo)
except KeyError:
dfs[y] = [periodo]
return dfs
def data_for_one_file(file, channel_name, dfs):
pol = polygram_from_pkl(file)
for t, w in pol.iter_window(WINDOW_SIZE, WINDOW_LAG):
eeg = w[channel_name]
ann = w["vigilance_state"]
if ann.probas.all() >0:
y = ann.values[0]
periodo = make_periodogram(eeg)
try:
dfs[y].append(periodo)
except KeyError:
dfs[y]=[periodo]
return dfs
from pyrem.time_series import Signal
from pyrem.polygram import Polygram
import pylab as pl
import pandas as pd
import numpy as np
# DATA_FILE_PATTERN=
#df = pd.read_csv("/tmp/telc4_res.csv")
#an = Annotation(df["pred"], 0.2, df["conf_preds"], name="prediction")
#pol1 = Polygram([an])
pol = polygram_from_pkl("/data/pyrem/Ellys/pkls/TelC_4.pkl")
#pol = pol1.merge(pol)
z = decompose_signal(pol["EEG_parietal_frontal"], [1,2,3,4,5,6])
z = z.merge(pol["EEG_parietal_frontal"])
z["16h45m":"16h55m"].show()
dwtp = decompose_signal(pol["EEG_parietal_frontal"], [6], keep_a=False)
t = dwtp[0] ** 2
N=25
tt = np.log10(np.convolve(t, np.ones((N,))/N,"same"))
t = Signal(tt,t.fs,name="Power in cD_6")
print t.duration
