def max2interval(peaks, cols, minrate=40, maxrate=200):
"""
Returns intervals from timesMax, after filtering possible artefacts based on rate range (rates in min^(-1))
if two peaks are nearer than minrate or further than maxrate, the algorithm try to recognize if there's a false true or a peak missing
return: an nparray (N,3) containing in order the time, the ibi and the label
peaks: an nparray (N,3) containing in order the timestamp, the height and the label of each peak
minrate: represents the minimum(in min^(-1)) bpm to detect
maxrate: represents the maximum(in min^(-1)) bpm to detect
"""
maxRR=60/minrate
minRR=60/maxrate
RR=[]
timeRR=[]
# algoritmo alternativo
# beat artifacts correction
##inizializzazioni
timed_lbls = selcol(peaks, cols, ["TIME", "LAB"])
peaks_time = selcol(peaks, cols, "TIME")
# print "TIME %d, LAB %d" % (len(timed_lbls[:,0]), len(timed_lbls[:,1]))
tprev=peaks_time[0]
tfalse=tprev
for i in range(1, len(peaks_time)):
tact=peaks_time[i]
RRcurr=tact-tprev
if (RRcurr<minRR): # troppo breve: falso picco?
tfalse=tact #aspetto (non aggiorno tact) e salvo il falso campione
elif RRcurr>maxRR: # troppo lungo ho perso un picco?
RRcurr=tact-tfalse # provo con l'ultimo falso picco se esiste
tprev=tact # dopo ripartiro' da questo picco
if RRcurr>minRR and RRcurr<maxRR: # tutto OK
RR.append(RRcurr) #aggiungo valore RR
timeRR.append(tact) #aggiungo istante temporale
tfalse=tact #aggiorno falso picco
tprev=tact #aggiorno tprev
#calculates the labels foreach time in timeRR
try:
labelRR = get_row_for_col(timed_lbls, timeRR)[:,1]
except IndexError as e:
print "NO LABELS: ", e.message
labelRR=np.array([])
pass
#the result contains a 2D array with the times and the ibi
return np.column_stack((timeRR, RR, labelRR)), ["TIME", "IBI", "LAB"]
def getPeaksIBI(signal, cols, peakDelta, s_type):
'''
get the peaks for the ibi calculation
return: an nparrays (N,3), containing in order the time (in s), the height of the peak and the lables and the column names
signal: signal as np.array (N,3) in which search the peaks (columns: timestamp, signal, labels)
peakDelta: minimum peak height
SAMP_F: the sampling frequency of signal
cols: the column labels
s_type=signal type: BVP or GSR
'''
timed_lbls = selcol(signal, cols, ["TIME", "LAB"])
t = selcol(signal, cols, "TIME")
signal_vals = selcol(signal ,cols, s_type)
maxp, minp = peakdet(signal_vals, peakDelta, t)
try:
new_lbls = get_row_for_col(timed_lbls, maxp[:,0])[:,1]
except IndexError as e:
print "IBI LABS: ", e.message
new_lbls = np.array([])
pass
cols_out=["TIME", "PEAK", "LAB"]
return np.column_stack((maxp, new_lbls)), cols_out
def extract_IBI_features(data, cols, windows, labels):
'''
exetract the features from an IBI with passed windows
return: np.array containing the features of each window, the new labels (removes the ones corresponding to empty windows)
data: the data from which extract the features, passed as a numpy array (N,2) with time and values in the two columns
windows: numpy array containing the start and the end point (in time) of every windows
labels: np.array containing the labels
cols: the column labels
'''
result = []
res_labels = labels
times = selcol(data, cols, "TIME")
values = selcol(data, cols, "IBI")
for tindex in xrange(len(windows)):
startt = windows[tindex][0]
endt = windows[tindex][1]
this_win = values[(times >= startt)&(times < endt)]
if this_win.size <= 1:
res_labels[tindex]=np.nan
else:
result.append(calculateHRVindexes(this_win))
res_labels=res_labels[~np.isnan(res_labels)]
return np.nan_to_num(result), res_labels
