def get_explicit_letter_intervals(
markers='./001_trial1_right_keyboard_2020-02-16-19-09-10-309.txt',
fname='./001_trial1_right_keyboard_OpenBCI-RAW-2020-02-16_18-59-08.txt',
channel=[1, 2, 3, 4]):
# load the data and find where they all are
channel.append(13)
channel1 = channel[:]
lr = load_raw.load_dta(markers, fname, channel)
key = lr[lr['keypressed'].notna()]
keypress_indices = [key.iloc[i].name for i in range(len(key))]
# the intervals should go from -350 till +500, it gets trimmed by chop_idx later which should be set to 250
intervals = [(i - 600, i + 600) for i in keypress_indices]
letters = key['keypressed'].values
return intervals, letters
def disp1(channel=[1, 2, 3, 4], save_as='spec1.png'):
lr = load_raw.load_dta(channel=channel + [13])
# trim data
keypressed = lr['keypressed']
start_idx = keypressed.first_valid_index()
stop_idx = keypressed.last_valid_index()
lr_trimmed = lr.truncate(before=start_idx - 750, after=stop_idx + 750)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# plot multiple channels
markings = np.array(
[only_keypressed.iloc[i].name for i in range(len(only_keypressed))])
labels = np.asarray(only_keypressed['keypressed'])
markings_labels = np.asarray([markings, labels]).T
ch_names = [
i for i in lr.columns.values
if i not in ('timestamp(ms)', 'keypressed')
]
eeg = np.asarray(
[display_methods.filter_signal(np.asarray(lr[i])) for i in ch_names])
fig = plt.figure(figsize=(10, int(5 * len(ch_names))),
dpi=80,
facecolor='w',
edgecolor='k')
for idx, ch in enumerate(eeg):
stop_plot_idx = start_idx + 10000
t, spec_freqs, spec_PSDperBin = get_spectral_content(
ch[start_idx:stop_plot_idx], SAMPLING_FREQ)
plot_specgram(spec_freqs,
spec_PSDperBin,
'channel = {}'.format(ch_names[idx]),
shift=0,
no_chan=len(ch_names),
i=idx + 1,
cmap='terrain')
# trim the markings and labels to only those that are within the specified range
markings_labels_trimmed = [(m, l) for m, l in markings_labels
if m < stop_plot_idx and m > start_idx]
for mark, label in markings_labels_trimmed:
plt.text((mark - start_idx) / 250,
.2,
label,
color='white',
fontsize=20)
plt.savefig(save_as)
def take_pictures(fname_dta=[
('./001_trial1_right_keyboard_2020-02-16-19-09-10-309.txt',
'./001_trial1_right_keyboard_OpenBCI-RAW-2020-02-16_18-59-08.txt')
],
channel=[1, 2, 3, 4],
figsize=(10, 12),
folder_name='pictures'):
channel1 = channel[:]
# make a folder
os.mkdir(folder_name) # throws exception if file already exists
# for each of the specified files load the data and take pictures
count1 = 0
for abc in fname_dta:
count1 += 1
markers = abc[0]
fname = abc[1]
# load the data, and get intervals information
lr = load_raw.load_dta(markers, fname, channel + [13])
explicit_intervals, letters = get_explicit_letter_intervals(
markers=markers, fname=fname, channel=channel)
count2 = 0
for i, j in zip(explicit_intervals, letters):
count2 += 1
channel = channel1[:]
try:
plot_ts_filtered(markers=markers,
fname=fname,
channel=channel,
figsize=figsize,
explicit_interval=i,
save_fig='./' + folder_name + '/' + j + '_' +
str(count1) + '_' + str(count2) + '_' +
markers[2:-4] + '.png',
disp=False)
except:
continue
def plot_ts_filtered(
markers='./001_trial1_right_keyboard_2020-02-16-19-09-10-309.txt',
fname='./001_trial1_right_keyboard_OpenBCI-RAW-2020-02-16_18-59-08.txt',
channel=[1, 2, 3, 4],
figsize=(8, 15),
time_interval=None,
explicit_interval=None,
save_fig=None,
disp=False):
"""
plot labelled time series data
time interval if specified must be a two-tuple with values (x,y) 0<x<y<1
explicit_interval is just indices (start, stop)
"""
# load labelled raw data
channel.append(13)
lr = load_raw.load_dta(markers, fname, channel)
# trim the data frame for plotting
keypressed = lr['keypressed']
try:
start_idx = keypressed.first_valid_index()
stop_idx = keypressed.last_valid_index()
# by default trim like this - to be rectified in following conditional statments if needed
lr_trimmed = lr.truncate(before=start_idx - 500, after=stop_idx + 500)
except:
start_idx = 0
stop_idx = 50000
lr_trimmed = lr.truncate(before=start_idx, after=stop_idx)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# trim more if user wants it to be trimmed
if time_interval:
start_idx_new = int(start_idx + time_interval[0] *
(stop_idx - start_idx))
stop_idx_new = int(start_idx + time_interval[1] *
(stop_idx - start_idx))
lr_trimmed = lr.truncate(before=start_idx_new, after=stop_idx_new)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# trim according to an explicit interval - specially used when saveing pngs
elif explicit_interval:
start_idx = explicit_interval[0]
stop_idx = explicit_interval[1]
lr_trimmed = lr.truncate(before=start_idx, after=stop_idx)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# plot the data
the_channels = [
i for i in lr.columns if i not in ('timestamp(ms)', 'keypressed')
]
count = 0
fig = plt.figure(figsize=figsize)
for channel_name in the_channels:
count += 1
ch = lr_trimmed[channel_name].values
timestamps = lr_trimmed['timestamp(ms)'].values
###
ch_keypress = only_keypressed[channel_name].values
ts_keypress = only_keypressed['timestamp(ms)'].values
np.unique(only_keypressed['keypressed'].values)
keys_pressed_unique = np.unique(only_keypressed['keypressed'].values)
ts_keypress = []
ch_i_keypress = []
labels = []
for letter in keys_pressed_unique:
labels.append('$' + letter + '$')
ch_i_keypress.append(only_keypressed.loc[
only_keypressed['keypressed'] == letter][channel_name].values)
ts_keypress.append(
only_keypressed.loc[only_keypressed['keypressed'] ==
letter]['timestamp(ms)'].values)
# make the plot - raw data
chop_idx = 250
plt.subplot(len(the_channels), 1, count)
print('channel_name', channel_name) #channel_name
plt.title(channel_name, fontsize=17)
# frequency filter
freq_filtered = filter_signal(ch)
plt.plot(np.arange(len(freq_filtered))[chop_idx:],
freq_filtered[chop_idx:],
label='frequency notch filter',
color='green',
alpha=.3)
# smooth the filtered with convolution
freq_filtered_conv = mmav2(freq_filtered, window=16)
freq_filtered_conv_abs = mmav2(np.abs(freq_filtered), window=10)
freq_filtered_rms_15 = rms(freq_filtered, 15)
freq_filtered_rms_5 = rms(freq_filtered, 5)
freq_filtered_rms_30 = rms(freq_filtered, 30)
freq_diff_smooth = mmav2(np.abs(np.diff(freq_filtered)), window=10)
plt.plot(np.arange(len(freq_filtered_conv))[chop_idx:],
freq_filtered_conv[chop_idx:],
label='conv',
alpha=.75)
plt.plot(np.arange(len(freq_filtered_conv_abs))[chop_idx:],
freq_filtered_conv_abs[chop_idx:],
label='conv abs',
alpha=.75)
plt.plot(np.arange(len(freq_filtered_rms_15))[chop_idx:],
freq_filtered_rms_15[chop_idx:],
label='rms 15',
alpha=.5)
plt.plot(np.arange(len(freq_filtered_rms_5))[chop_idx:],
freq_filtered_rms_5[chop_idx:],
label='rms 5',
alpha=.5)
plt.plot(np.arange(len(freq_filtered_rms_30))[chop_idx:],
freq_filtered_rms_30[chop_idx:],
label='rms 30',
alpha=.6)
plt.plot(np.arange(len(freq_diff_smooth))[chop_idx:],
freq_diff_smooth[chop_idx:],
label='freq abs difference',
color='red',
alpha=.35)
l = list(lr_trimmed['timestamp(ms)'].values)
for x, y, lab in zip(ts_keypress, ch_i_keypress, labels):
x_new = [l.index(i) for i in x]
plt.plot(x_new,
np.ones(len(y)) * 75.0,
marker=lab,
color='red',
linestyle='None')
plt.legend(fontsize=12, loc=1, framealpha=.1)
# set the y axis max min if i'm taking pictures
if explicit_interval:
x1, x2, y1, y2 = plt.axis()
plt.axis((x1, x2, -65, 175))
if save_fig: plt.savefig(save_fig)
if disp: plt.show()
plt.close(fig)
def plot_ts_2(
markers='./001_trial1_right_keyboard_2020-02-16-19-09-10-309.txt',
fname='./001_trial1_right_keyboard_OpenBCI-RAW-2020-02-16_18-59-08.txt',
channel=[1, 2, 3, 4],
figsize=(8, 15),
time_interval=None): #no_dta_pts=None
"""
plot labelled time series data
time interval if specified must be a two-tuple with values (x,y) 0<x<y<1
"""
# load labelled raw data
channel.append(13)
lr = load_raw.load_dta(markers, fname, channel)
# trim the data frame for plotting
keypressed = lr['keypressed']
start_idx = keypressed.first_valid_index()
stop_idx = keypressed.last_valid_index()
lr_trimmed = lr.truncate(before=start_idx - 500, after=stop_idx + 500)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# trim more if user wants it to be trimmed
if time_interval:
start_idx_new = int(start_idx + time_interval[0] *
(stop_idx - start_idx))
stop_idx_new = int(start_idx + time_interval[1] *
(stop_idx - start_idx))
lr_trimmed = lr.truncate(before=start_idx_new, after=stop_idx_new)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# plot the data
channels = [
i for i in lr.columns if i not in ('timestamp(ms)', 'keypressed')
]
count = 0
plt.figure(figsize=figsize)
for channel_name in channels:
count += 1
ch = lr_trimmed[channel_name].values
timestamps = lr_trimmed['timestamp(ms)'].values
###
ch_keypress = only_keypressed[channel_name].values
ts_keypress = only_keypressed['timestamp(ms)'].values
np.unique(only_keypressed['keypressed'].values)
keys_pressed_unique = np.unique(only_keypressed['keypressed'].values)
ts_keypress = []
ch_i_keypress = []
labels = []
for letter in keys_pressed_unique:
labels.append('$' + letter + '$')
ch_i_keypress.append(only_keypressed.loc[
only_keypressed['keypressed'] == letter][channel_name].values)
ts_keypress.append(
only_keypressed.loc[only_keypressed['keypressed'] ==
letter]['timestamp(ms)'].values)
# make the plot - raw data
plt.subplot(len(channels), 1, count)
plt.title(channel_name, fontsize=15)
plt.plot(ch, label='raw')
# filter the signal and plot smoothed
filt_conv_1 = conv_1(ch, 100)
plt.plot(np.arange(len(filt_conv_1))[100:-100],
filt_conv_1[100:-100],
label='conv filter')
# frequency filter
freq_filtered = filter_signal(ch)
baseline = np.mean(ch)
plt.plot(np.arange(len(freq_filtered))[150:],
freq_filtered[150:] +
np.ones(len(freq_filtered) - 150) * baseline,
label='frequency filter',
alpha=0.3)
l = list(lr_trimmed['timestamp(ms)'].values)
for x, y, lab in zip(ts_keypress, ch_i_keypress, labels):
x_new = [l.index(i) for i in x]
plt.plot(x_new,
y + np.ones(len(y)) * 300,
marker=lab,
color='red',
linestyle='None')
plt.legend()
#plt.savefig('djjd.png')
plt.show()
def plot_ts(
markers='./001_trial1_right_keyboard_2020-02-16-19-09-10-309.txt',
fname='./001_trial1_right_keyboard_OpenBCI-RAW-2020-02-16_18-59-08.txt',
channel=[1, 2, 3, 4],
figsize=(8, 15),
time_interval=None): #no_dta_pts=None
"""
plot labelled time series data
time interval if specified must be a two-tuple with values (x,y) 0<x<y<1
"""
# load labelled raw data
channel.append(13)
lr = load_raw.load_dta(markers, fname, channel)
# trim the data frame for plotting
keypressed = lr['keypressed']
start_idx = keypressed.first_valid_index()
stop_idx = keypressed.last_valid_index()
lr_trimmed = lr.truncate(before=start_idx - 500, after=stop_idx + 500)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# trim more if user wants it to be trimmed
if time_interval:
start_idx_new = int(start_idx + time_interval[0] *
(stop_idx - start_idx))
stop_idx_new = int(start_idx + time_interval[1] *
(stop_idx - start_idx))
lr_trimmed = lr.truncate(before=start_idx_new, after=stop_idx_new)
only_keypressed = lr_trimmed[lr_trimmed['keypressed'].notna()]
# plot the data
channels = [
i for i in lr.columns if i not in ('timestamp(ms)', 'keypressed')
]
count = 0
plt.figure(figsize=figsize)
for channel_name in channels:
count += 1
ch = lr_trimmed[channel_name].values
timestamps = lr_trimmed['timestamp(ms)'].values
###
ch_keypress = only_keypressed[channel_name].values
ts_keypress = only_keypressed['timestamp(ms)'].values
np.unique(only_keypressed['keypressed'].values)
keys_pressed_unique = np.unique(only_keypressed['keypressed'].values)
ts_keypress = []
ch_i_keypress = []
labels = []
for letter in keys_pressed_unique:
labels.append('$' + letter + '$')
ch_i_keypress.append(only_keypressed.loc[
only_keypressed['keypressed'] == letter][channel_name].values)
ts_keypress.append(
only_keypressed.loc[only_keypressed['keypressed'] ==
letter]['timestamp(ms)'].values)
# make the plot - raw data
plt.subplot(len(channels), 1, count)
plt.title(channel_name, fontsize=15)
plt.plot(timestamps, ch)
for x, y, lab in zip(ts_keypress, ch_i_keypress, labels):
plt.plot(x,
y + np.ones(len(y)) * 300,
marker=lab,
color='red',
linestyle='None')
plt.savefig('asdf.png')
plt.show()