def events_find_events(self, raw, **param):
"""
some how calls <mne.find_events()>
input:
raw obj,
e.g. parameters:
{'event_id': 40, 'and_mask': 255,
'events': {'consecutive': True, 'output':'steps','stim_channel': 'STI 014',
'min_duration':0.002,'shortest_event': 2,'mask': 0}
}
return:
pandas data-frame with epoch event structure for stimulus or response channel
id : event id
offset : np array with TSL event code offset
onset : np array with TSL event code onset
counts : number of events
bads : np.array with index of bad events
#=> RESPONSE MATCHING
#rt_type : NAN => np array with values MISSED=0 ,TOEARLY=1,WRONG=2,HIT=3
#rt : NAN => np array with reaction time [TSL]
#rt_onset: NAN => np array with response onset [TSL]
#rt_id : NAN => np array with response key / id
dict() with event structure for stimulus or response channel
sfreq : sampling frequency => raw.info['sfreq']
duration : {mean,min,max} in TSL
system_delay_is_applied : True/False
--> if true <system_delay_ms> converted to TSLs and added to the TSLs in onset,offset
(TSL => timeslices , your samples)
"""
if raw is None:
print "ERROR in <get_event_structure: No raw obj \n"
return None
#---
import pandas as pd
df = pd.DataFrame(columns=self.data_frame_stimulus_cols)
#---
events = param['events'].copy()
events['output'] = 'step'
ev = mne.find_events(raw, **events) #-- return int64
#--- apply and mask e.g. 255 get the first 8 bits in Trigger channel
if param['and_mask']:
ev[:, 1:] = np.bitwise_and(ev[:, 1:], param['and_mask'])
ev[:, 2:] = np.bitwise_and(ev[:, 2:], param['and_mask'])
#--- search and return only events with event_id in event_id-array
#--- shape =>(trials,3) => [trial idx ,tsl-onset/ tsl-offset,event id]
#--- split this in onset offset
ev_onset = np.squeeze(ev[np.where(ev[:, 2]), :]) # > 0
ev_offset = np.squeeze(ev[np.where(ev[:, 1]), :])
if param['event_id']:
ev_id = jumeg_base.str_range_to_numpy(param['event_id'],
exclude_zero=True)
#ev_onset = np.squeeze( ev[np.where( np.in1d( ev[:,2],ev_id ) ), :])
#ev_offset = np.squeeze( ev[np.where( np.in1d( ev[:,1],ev_id ) ), :])
ev_id_idx = np.squeeze(np.where(np.in1d(ev_onset[:, 2], ev_id)))
ev_onset = ev_onset[ev_id_idx, :]
ev_offset = ev_offset[ev_id_idx, :]
if (ev_onset.size == 0):
return df
#--- apply system delay if is defined e.g. auditory take`s 20ms to subjects ears
if param['system_delay_ms']:
system_delay_tsl = raw.time_as_index(param['system_delay_ms'] /
1000) # calc in sec
ev_onset[:, 0] += system_delay_tsl
ev_offset[:, 0] += system_delay_tsl
system_delay_is_applied = True
else:
system_delay_is_applied = False
#-- avoid invalid index/dimension error if last offset is none
div = ev_offset[:, 0] - ev_onset[:ev_offset[:, 0].size, 0]
df['id'] = ev_onset[:, 2]
df['onset'] = ev_onset[:, 0]
df['offset'] = np.zeros(ev_onset[:, 0].size, dtype=np.long)
df['offset'][:ev_offset[:, 0].size] = ev_offset[:, 0]
return df, dict({
'sfreq': raw.info['sfreq'],
'duration': {
'mean': np.rint(div.mean()),
'min': div.min(),
'max': div.max()
},
'system_delay_is_applied': system_delay_is_applied
})
def events_response_matching(self,
raw,
stim_df=None,
resp_df=None,
**param):
"""
matching correct responses with respect to <stimulus channel> <output type> (onset,offset)
input:
stim_df = <stimulus channel data frame>
res_df = <response channel data frame>
param : parameter dict
e.g.:{ 'response':{'counts':1,'window':[0,0.5],'event_id':2,'include_early_ids':[1,4],'events':{'output':'onset'} },
'stimulus':{'events':{'output'.'onset'}} }
return:
<stimulus data frame> pandas DataFrame obj
with added cols
rt_type : response type MISSED,EARLY,WRONG,HIT
rt : response time as tsl (RT)
rt_id : event id button press code
rt_onset : response onset [tsl]
rt_offset : response offset [tsl]
rt_index : index in response onset / offset
rt_counts : number of responses in <response time window>
bads : flag for bad epochs; e.g. later use in combination with ecg/eog events
"""
#--- ck errors
err_msg = None
if raw is None:
err_msg = "ERROR in <apply_response_matching> : no raw obj \n"
if (stim_df is None):
err_msg += "\nERROR no Stimulus-Data-Frame obj. provided\n"
if (resp_df is None):
err_msg += "\nERROR no Response-Data-Frame obj. provided\n"
if (param is None):
err_msg += "\nERROR no Parameter obj. provided\n"
#--- ck RT window range
if (param['response']['window'][0] >= param['response']['window'][1]):
err_msg += "ERROR in parameter response windows\n"
if err_msg:
print "ERROR in <apply_response_matching>\n" + err_msg + "\n\n"
return None
#--- extend stimulus data frame
for x in self.data_frame_response_cols:
stim_df[x] = 0 #np.NaN
#--- convert rt window [ms] into tsl
(r_window_tsl_start,
r_window_tsl_end) = raw.time_as_index(param['response']['window'])
#--- get respose code -> event_id [int or string] as np array
r_event_id = jumeg_base.str_range_to_numpy(
param['response']['event_id'])
#--- ck if any toearly-id is defined, returns None if not
r_event_id_toearly = jumeg_base.str_range_to_numpy(
param['response']['include_early_ids'])
#--- get output type: onset or offset
stim_output_type = param['stimulus']['events']['output']
resp_output_type = param['response']['events']['output']
#--- loop for all stim events
ridx = 0
#--- get rt important part of respose df
RESP_TSLs = resp_df[resp_output_type]
for idx in stim_df.index:
st_tsl_onset = stim_df[stim_output_type][idx]
st_window_tsl0 = stim_df[stim_output_type][idx] + r_window_tsl_start
st_window_tsl1 = stim_df[stim_output_type][idx] + r_window_tsl_end
# st_id = stim_df['id'][idx]
#--- check for to TOEARLY responses
toearly_tsl0 = st_tsl_onset
toearly_tsl1 = st_window_tsl0
#--- look for part of response dataframe ...
res_df_early = resp_df[(toearly_tsl0 <= RESP_TSLs)
& (RESP_TSLs < toearly_tsl1)]
if res_df_early.index.size > 0:
if not res_df_early.isin(r_event_id_toearly).all:
ridx = res_df_early.index[0]
stim_df.rt_counts[idx] = res_df_in.index.size
stim_df.rt_onset[idx] = resp_df.onset[ridx]
stim_df.rt_offset[idx] = resp_df.offset[ridx]
stim_df.rt_id[idx] = resp_df.id[ridx]
stim_df.rt_index[idx] = ridx
stim_df.rt_type[idx] = self.idx_toearly
continue
#--- find index of responses from window-start till end of res_event_type array [e.g. onset / offset]
# res_df_in = resp_df[ ( st_window_tsl0 <= RESP_TSLs ) & ( RESP_TSLs <= st_window_tsl1) ]
resp_df_in_idx = resp_df[(st_window_tsl0 <= RESP_TSLs)
& (RESP_TSLs <= st_window_tsl1)].index
#--- MISSED response
if not np.any(resp_df_in_idx):
continue
#--- WRONG or HIT;
# if res_df_in.index.size > 0 :
# ridx = res_df_in.index[0]
if resp_df_in_idx.size > 0:
ridx = resp_df_in_idx[0]
stim_df.rt_counts[idx] = resp_df_in_idx.size
stim_df.rt_onset[idx] = resp_df.onset[ridx]
stim_df.rt_offset[idx] = resp_df.offset[ridx]
stim_df.rt_type[idx] = self.idx_wrong
stim_df.rt_id[idx] = resp_df.id[ridx]
#stim_df.rt_id[idx] = resp_df.id[resp_df_in_idx].max()
#stim_df.rt_id[idx] = np.bitwise_and(resp_df.id[resp_df_in_idx], 72).max()
stim_df.rt_index[idx] = ridx
#--- HIT; ck number of responses; ck pressed buttons; wrong if more than count
if (stim_df.rt_counts[idx] <= param['response']['counts']):
#if np.all( res_df_in.id.isin( r_event_id ) ) :
if np.all(resp_df.id[resp_df_in_idx].isin(r_event_id)):
stim_df.rt_type[idx] = self.idx_hit
# if (stim_df.rt_id[idx] <1) and(stim_df.rt_type[idx] == self.idx_wrong):
# print"!!ERROR"
# print resp_df_in_idx
# print resp_df.onset[resp_df_in_idx]
# print resp_df.id[resp_df_in_idx]
# assert "erroe rt"
#--- MISSED response
else:
continue
#--- calc reaction time (rt in tsl)
if stim_output_type == 'offset':
sto = stim_df.offset
else:
sto = stim_df.onset
if resp_output_type == 'offset':
rto = stim_df.rt_offset
else:
rto = stim_df.rt_onset
stim_df.rt = rto - sto
stim_df.rt[(stim_df.rt_type == self.idx_missed)] = 0
#---
if self.verbose:
for kidx in range(len(self.rt_type_list)):
print "\n\n---> Stimulus DataFrame Type: " + self.rt_type_list[
kidx]
ddf = stim_df[stim_df.rt_type == kidx]
if ddf.index.size > 0:
print ddf.describe()
else:
print "---> EMPTY"
print "---------------------------"
print "\n\n"
#import sys
#char = sys.stdin.read(1)
return stim_df
def events_find_events(self,raw,**param):
"""
some how calls <mne.find_events()>
input:
raw obj,
e.g. parameters:
{'event_id': 40, 'and_mask': 255,
'events': {'consecutive': True, 'output':'steps','stim_channel': 'STI 014',
'min_duration':0.002,'shortest_event': 2,'mask': 0}
}
return:
pandas data-frame with epoch event structure for stimulus or response channel
id : event id
offset : np array with TSL event code offset
onset : np array with TSL event code onset
counts : number of events
bads : np.array with index of bad events
#=> RESPONSE MATCHING
#rt_type : NAN => np array with values MISSED=0 ,TOEARLY=1,WRONG=2,HIT=3
#rt : NAN => np array with reaction time [TSL]
#rt_onset: NAN => np array with response onset [TSL]
#rt_id : NAN => np array with response key / id
dict() with event structure for stimulus or response channel
sfreq : sampling frequency => raw.info['sfreq']
duration : {mean,min,max} in TSL
system_delay_is_applied : True/False
--> if true <system_delay_ms> converted to TSLs and added to the TSLs in onset,offset
(TSL => timeslices , your samples)
"""
if raw is None:
print "ERROR in <get_event_structure: No raw obj \n"
return None
#---
import pandas as pd
df = pd.DataFrame(columns = self.data_frame_stimulus_cols)
#---
events = param['events'].copy()
events['output'] = 'step'
ev = mne.find_events(raw, **events)
#--- apply and mask e.g. 255 get the first 8 bits in Trigger channel
if param['and_mask']:
ev[:, 1:] = np.bitwise_and(ev[:, 1:], param['and_mask'])
ev[:, 2:] = np.bitwise_and(ev[:, 2:], param['and_mask'])
#--- search and return only events with event_id in event_id-array
#--- shape =>(trials,3) => [trial idx ,tsl-onset/ tsl-offset,event id]
#--- split this in onset offset
if param['event_id']:
ev_id = jumeg_base.str_range_to_numpy(param['event_id'])
ev_onset = np.squeeze( ev[np.where( np.in1d( ev[:,2],ev_id ) ), :])
ev_offset = np.squeeze( ev[np.where( np.in1d( ev[:,1],ev_id ) ), :])
#--- get all ids e.g. response for later response matching
else:
ev_onset = ev[0::2,]
ev_offset = ev[1::2,]
if ( ev_onset.size == 0 ):
return df
#--- apply system delay if is defined e.g. auditory take`s 20ms to subjects ears
if param['system_delay_ms']:
system_delay_tsl = raw.time_as_index( param['system_delay_ms']/1000 ) # calc in sec
ev_onset[:, 0] += system_delay_tsl
ev_offset[:, 0]+= system_delay_tsl
system_delay_is_applied = True
else:
system_delay_is_applied = False
div = ev_offset[:,0] - ev_onset[:,0]
df['id'] = ev_onset[:,2]
df['onset'] = ev_onset[:,0]
df['offset'] = ev_offset[:,0]
return df,dict( {
'sfreq' : raw.info['sfreq'],
'duration' :{'mean':np.rint(div.mean()),'min':div.min(),'max':div.max()},
'system_delay_is_applied' : system_delay_is_applied
} )
def events_response_matching(self,raw,stim_df=None,resp_df=None, **param ):
"""
matching correct responses with respect to <stimulus channel> <output type> (onset,offset)
input:
stim_df = <stimulus channel data frame>
res_df = <response channel data frame>
param : parameter dict
e.g.:{ 'response':{'counts':1,'window':[0,0.5],'event_id':2,'include_early_ids':[1,4],'events':{'output':'onset'} },
'stimulus':{'events':{'output'.'onset'}} }
return:
<stimulus data frame> pandas DataFrame obj
with added cols
rt_type : response type MISSED,EARLY,WRONG,HIT
rt : response time as tsl (RT)
rt_id : event id button press code
rt_onset : response onset [tsl]
rt_offset : response offset [tsl]
rt_index : index in response onset / offset
rt_counts : number of responses in <response time window>
bads : flag for bad epochs; e.g. later use in combination with ecg/eog events
"""
#--- ck errors
err_msg = None
if raw is None:
err_msg = "ERROR in <apply_response_matching> : no raw obj \n"
if (stim_df is None):
err_msg +="\nERROR no Stimulus-Data-Frame obj. provided\n"
if (resp_df is None):
err_msg +="\nERROR no Response-Data-Frame obj. provided\n"
if (param is None):
err_msg +="\nERROR no Parameter obj. provided\n"
#--- ck RT window range
if ( param['response']['window'][0] >= param['response']['window'][1] ):
err_msg += "ERROR in parameter response windows\n"
if err_msg:
print "ERROR in <apply_response_matching>\n" + err_msg +"\n\n"
return None
#--- extend stimulus data frame
for x in self.data_frame_response_cols :
stim_df[x]= 0 #np.NaN
#--- convert rt window [ms] into tsl
(r_window_tsl_start, r_window_tsl_end ) = raw.time_as_index( param['response']['window'] );
#--- get respose code -> event_id [int or string] as np array
r_event_id = jumeg_base.str_range_to_numpy( param['response']['event_id'] )
#--- ck if any toearly-id is defined, returns None if not
r_event_id_toearly = jumeg_base.str_range_to_numpy( param['response']['include_early_ids'] )
#--- get output type: onset or offset
stim_output_type = param['stimulus']['events']['output']
resp_output_type = param['response']['events']['output']
#--- loop for all stim events
ridx = 0
#--- get rt important part of respose df
RESP_TSLs = resp_df[resp_output_type]
for idx in stim_df.index :
st_tsl_onset = stim_df[stim_output_type][idx]
st_window_tsl0 = stim_df[stim_output_type][idx] + r_window_tsl_start
st_window_tsl1 = stim_df[stim_output_type][idx] + r_window_tsl_end
st_id = stim_df['id'][idx]
#--- check for to TOEARLY responses
toearly_tsl0 = st_tsl_onset
toearly_tsl1 = st_window_tsl0
#--- look for part of response dataframe ...
res_df_early = resp_df[(toearly_tsl0 <= RESP_TSLs ) & ( RESP_TSLs < toearly_tsl1 )]
if res_df_early.index.size > 0 :
if not res_df_early.isin( r_event_id_toearly ).all :
ridx = res_df_early.index[0]
stim_df.rt_counts[idx] = res_df_in.index.size
stim_df.rt_onset[idx] = resp_df.onset[ridx]
stim_df.rt_offset[idx] = resp_df.offset[ridx]
stim_df.rt_id[idx] = resp_df.id[ridx]
stim_df.rt_index[idx] = ridx
stim_df.rt_type[idx] = self.idx_toearly
continue
#--- find index of responses from window-start till end of res_event_type array [e.g. onset / offset]
res_df_in = resp_df[ ( st_window_tsl0 <= RESP_TSLs ) & ( RESP_TSLs <= st_window_tsl1) ]
#--- MISSED response
if res_df_in.empty : continue
#--- WRONG or HIT;
if res_df_in.index.size > 0 :
ridx = res_df_in.index[0]
stim_df.rt_counts[idx] = res_df_in.index.size
stim_df.rt_onset[idx] = resp_df.onset[ridx]
stim_df.rt_offset[idx] = resp_df.offset[ridx]
stim_df.rt_type[idx] = self.idx_wrong
stim_df.rt_id[idx] = resp_df.id[ridx]
stim_df.rt_index[idx] = ridx
#--- HIT; ck number of responses; ck pressed buttons; wrong if more than count
if ( stim_df.rt_counts[idx] <= param['response']['counts'] ):
if res_df_in.id.isin( r_event_id ).all :
stim_df.rt_type[idx] = self.idx_hit
#--- calc reaction time (rt in tsl)
if stim_output_type == 'offset' :
sto = stim_df.offset
else:
sto = stim_df.onset
if resp_output_type == 'offset' :
rto = stim_df.rt_offset
else:
rto = stim_df.rt_onset
stim_df.rt = rto - sto
#---
if self.verbose:
for kidx in range( len( self.rt_type_list ) ) :
print "\n\n---> Stimulus DataFrame Type: " + self.rt_type_list[kidx]
ddf = stim_df[ stim_df.rt_type == kidx ]
if ddf.index.size > 0 :
print ddf.describe()
else :
print "---> EMPTY"
print"---------------------------"
print "\n\n"
#import sys
#char = sys.stdin.read(1)
return stim_df
def events_response_matching(self,raw,stim_df=None,resp_df=None, **param ):
"""
matching correct responses with respect to <stimulus channel> <output type> (onset,offset)
input:
stim_df = <stimulus channel data frame>
res_df = <response channel data frame>
param : parameter dict
e.g.:{ 'response':{'counts':1,'window':[0,0.5],'event_id':2,'include_early_ids':[1,4],'events':{'output':'onset'} },
'stimulus':{'events':{'output'.'onset'}} }
return:
<stimulus data frame> pandas DataFrame obj
with added cols
rt_type : response type MISSED,EARLY,WRONG,HIT
rt : response time as tsl (RT)
rt_id : event id button press code
rt_onset : response onset [tsl]
rt_offset : response offset [tsl]
rt_index : index in response onset / offset
rt_counts : number of responses in <response time window>
bads : flag for bad epochs; e.g. later use in combination with ecg/eog events
"""
#--- ck errors
err_msg = None
if raw is None:
err_msg = "ERROR in <apply_response_matching> : no raw obj \n"
if (stim_df is None):
err_msg +="\nERROR no Stimulus-Data-Frame obj. provided\n"
if (resp_df is None):
err_msg +="\nERROR no Response-Data-Frame obj. provided\n"
if (param is None):
err_msg +="\nERROR no Parameter obj. provided\n"
#--- ck RT window range
if ( param['response']['window'][0] >= param['response']['window'][1] ):
err_msg += "ERROR in parameter response windows\n"
if err_msg:
print "ERROR in <apply_response_matching>\n" + err_msg +"\n\n"
return None
#--- extend stimulus data frame
for x in self.data_frame_response_cols :
stim_df[x]= 0 #np.NaN
#--- convert rt window [ms] into tsl
(r_window_tsl_start, r_window_tsl_end ) = raw.time_as_index( param['response']['window'] );
#--- get respose code -> event_id [int or string] as np array
r_event_id = jumeg_base.str_range_to_numpy( param['response']['event_id'] )
#--- ck if any toearly-id is defined, returns None if not
r_event_id_toearly = jumeg_base.str_range_to_numpy( param['response']['include_early_ids'] )
#--- get output type: onset or offset
stim_output_type = param['stimulus']['events']['output']
resp_output_type = param['response']['events']['output']
#--- loop for all stim events
ridx = 0
#--- get rt important part of respose df
RESP_TSLs = resp_df[resp_output_type]
for idx in stim_df.index :
st_tsl_onset = stim_df[stim_output_type][idx]
st_window_tsl0 = stim_df[stim_output_type][idx] + r_window_tsl_start
st_window_tsl1 = stim_df[stim_output_type][idx] + r_window_tsl_end
=======
ev = mne.find_events(raw, **events) #-- return int64
>>>>>>> fb_devel_03082015
#--- apply and mask e.g. 255 get the first 8 bits in Trigger channel
if param['and_mask']:
ev[:, 1:] = np.bitwise_and(ev[:, 1:], param['and_mask'])
ev[:, 2:] = np.bitwise_and(ev[:, 2:], param['and_mask'])
#--- search and return only events with event_id in event_id-array
#--- shape =>(trials,3) => [trial idx ,tsl-onset/ tsl-offset,event id]
#--- split this in onset offset
<<<<<<< HEAD
if param['event_id']:
ev_id = jumeg_base.str_range_to_numpy(param['event_id'])
ev_onset = np.squeeze( ev[np.where( np.in1d( ev[:,2],ev_id ) ), :])
ev_offset = np.squeeze( ev[np.where( np.in1d( ev[:,1],ev_id ) ), :])
#--- get all ids e.g. response for later response matching
else:
ev_onset = ev[0::2,]
ev_offset = ev[1::2,]
=======
ev_onset = np.squeeze( ev[np.where( ev[:,2] ),:]) # > 0
ev_offset = np.squeeze( ev[np.where( ev[:,1] ),:])
if param['event_id']:
ev_id = jumeg_base.str_range_to_numpy(param['event_id'],exclude_zero=True)