def pitjb_online_parser(v, p=None):
if p is None:
p = get_default_params()
v_peak = p.online_vthresh
peaks,ends = [],[]
pe = 0,0 # peaktime, sacendtime tuple which we'll use
# XXX: it's slow to iterate over the velocity like this, but it's the
# easiest way to have code that does the same thing as what our
# online-parser currently does
for i in range(len(v)):
if i == pe[1] - p.framelag:
ends.append(pe[1])
peaks.append(pe[0])
if v[i] >p.online_vthresh:
#print "Got higher than v_thresh"
if v[i] > v_peak:
v_peak = v[i]
peaktime = i
sacendtime = peaktime + timeleft_from_vpeak(v_peak,
p.sacslope,p.sacintercept)
sacendtime = int(sacendtime)
pe = peaktime,sacendtime
# communicate the predicted fixation position to visionegg display loop
# just put a random position for now
#fix_position[:] = np.random.rand() * 800, np.random.rand() * 600
#fix_position[:] = rbuf[i-1]
#print "updating v_peak", v[i], " \t curtime:", i, "end time: ",sacendtime
#peaks.append(i)
else:
# we're slower than threshold again, must be in a fixation epoch
v_peak = p.online_vthresh
return np.array(peaks),np.array(ends)
def pitjb_online_parser(v, p=None):
if p is None:
p = get_default_params()
v_peak = p.online_vthresh
peaks, ends = [], []
pe = 0, 0 # peaktime, sacendtime tuple which we'll use
# XXX: it's slow to iterate over the velocity like this, but it's the
# easiest way to have code that does the same thing as what our
# online-parser currently does
for i in range(len(v)):
if i == pe[1] - p.framelag:
ends.append(pe[1])
peaks.append(pe[0])
if v[i] > p.online_vthresh:
#print "Got higher than v_thresh"
if v[i] > v_peak:
v_peak = v[i]
peaktime = i
sacendtime = peaktime + timeleft_from_vpeak(
v_peak, p.sacslope, p.sacintercept)
sacendtime = int(sacendtime)
pe = peaktime, sacendtime
# communicate the predicted fixation position to visionegg display loop
# just put a random position for now
#fix_position[:] = np.random.rand() * 800, np.random.rand() * 600
#fix_position[:] = rbuf[i-1]
#print "updating v_peak", v[i], " \t curtime:", i, "end time: ",sacendtime
#peaks.append(i)
else:
# we're slower than threshold again, must be in a fixation epoch
v_peak = p.online_vthresh
return np.array(peaks), np.array(ends)
def pitjb_online_parser_fast(v, p=None):
"""Online parser which extracts saccades based on peak velocity that is
above some treshold.
v : array
velocity in *pixels* per sample
XXX: not working at the moment
"""
from Dimstim.multipylink import timeleft_from_vpeak
if p is None:
p = get_default_params()
v_over_thresh, = np.where(v > p.online_vthresh)
tleft = timeleft_from_vpeak(v[v_over_thresh],
p.sacslope,p.sacintercept).astype(np.int)
ends = v_over_thresh+tleft
# is the sample larger that all other velocities before the predicted end
# time
# XXX: this needs to be modified by our framelag / framerate, see
# Dimstim.FixBarsAsync for details , in particular lines that include
# either "ttl < 10 and ttl > 4" or "ttl < 5 and ttl >= 0"
# this won't work - we'll get a ragged array depending on when stuff ends
#mask = v[v_over_thresh] > v[v_over_thresh+np.arange(ends)].all()
mask = [(v[b] > v[b+1:e]).all() for b,e in zip(v_over_thresh, ends)]
mask = np.array(mask)
# if the mask condition is false, that means we would have avoided this
# sample, but if it's true, we still may have avoided this sample, if
# there's a sample before it that caused us to make a prediction further
# into the past.
print(mask)
print(v_over_thresh)
if mask.sum() > 1:
# now every peak thats at mask=True is the maximum until the end of its
# predicted saccade. If there are any other peaks before this end, we
# will ignore them (unless they occur after we've gone below
# threshold), in which case we'll re-predict a new saccade. How do we
# account for this in the code? what a headache!
mask2 = ends[mask][:-1] < v_over_thresh[mask][1:]
mask[1:][~mask2] = False
print(mask2)
return v_over_thresh[mask]
def pitjb_online_parser_fast(v, p=None):
"""Online parser which extracts saccades based on peak velocity that is
above some treshold.
v : array
velocity in *pixels* per sample
XXX: not working at the moment
"""
from Dimstim.multipylink import timeleft_from_vpeak
if p is None:
p = get_default_params()
v_over_thresh, = np.where(v > p.online_vthresh)
tleft = timeleft_from_vpeak(v[v_over_thresh], p.sacslope,
p.sacintercept).astype(np.int)
ends = v_over_thresh + tleft
# is the sample larger that all other velocities before the predicted end
# time
# XXX: this needs to be modified by our framelag / framerate, see
# Dimstim.FixBarsAsync for details , in particular lines that include
# either "ttl < 10 and ttl > 4" or "ttl < 5 and ttl >= 0"
# this won't work - we'll get a ragged array depending on when stuff ends
#mask = v[v_over_thresh] > v[v_over_thresh+np.arange(ends)].all()
mask = [(v[b] > v[b + 1:e]).all() for b, e in zip(v_over_thresh, ends)]
mask = np.array(mask)
# if the mask condition is false, that means we would have avoided this
# sample, but if it's true, we still may have avoided this sample, if
# there's a sample before it that caused us to make a prediction further
# into the past.
print(mask)
print(v_over_thresh)
if mask.sum() > 1:
# now every peak thats at mask=True is the maximum until the end of its
# predicted saccade. If there are any other peaks before this end, we
# will ignore them (unless they occur after we've gone below
# threshold), in which case we'll re-predict a new saccade. How do we
# account for this in the code? what a headache!
mask2 = ends[mask][:-1] < v_over_thresh[mask][1:]
mask[1:][~mask2] = False
print(mask2)
return v_over_thresh[mask]
