def emgsim():
# sampling rate of the simulated EMG data
fs = 2000
# update rate of the generated data
update_rate = 20
# gain to use in noise generation
gain = 0.25
# number of seconds of data the oscilloscope shows
osc_view_time = 5
samp_per_input = int(fs / update_rate)
pipeline = Pipeline([
# get keyboard inputs of past second
Windower(update_rate),
# take mean over last second and apply a gain
Callable(lambda x: np.mean(x, axis=1, keepdims=True)),
# generate noise with amplitude of previous output
Callable(lambda x, k: gain * x * np.random.randn(x.shape[0], k),
func_args=(samp_per_input, )),
# window for pretty display in oscilloscope
Windower(osc_view_time * update_rate * samp_per_input),
])
dev = Keyboard(rate=update_rate, keys=list('wasd'))
run(dev, pipeline)
def make_pipeline(self, gain):
b, a = butter(4, (10 / 2000. / 2., 900 / 2000. / 2.), 'bandpass')
pipeline = Pipeline([
Windower(250),
Filter(b, a=a, overlap=150), # overlap = winsize - read_rate
Callable(lambda x: x[:, 0].reshape(-1, 1)), # Downsampling
Callable(lambda x: gain * x)
])
return pipeline
def mouse():
dev = Mouse(rate=20)
pipeline = Pipeline([
# just for scaling the input since it's in pixels
Callable(lambda x: x / 100),
# window to show in the oscilloscope
Windower(40)
])
run(dev, pipeline)
def keystick():
dev = Keyboard(rate=20, keys=list('wasd'))
pipeline = Pipeline([
# window to average over
Windower(10),
# mean along rows
Callable(lambda x: np.mean(x, axis=1, keepdims=True)),
# window to show in the oscilloscope
Windower(60)
])
run(dev, pipeline)
def make_mav_norm_pipeline(self):
windower = Windower(int(self.rate * self.win_size_mav))
if self.filter:
filter = self.make_filter(win_size=self.win_size_mav)
fe = FeatureExtractor([('MAV', MeanAbsoluteValue())])
mmsc = MinMaxScaler(min_=self.c_min, max_=self.c_max)
# Calibrated MAV is non-negative, but we allow values larger than 1
nonneg = Callable(lambda x: np.clip(x, 0.0, None))
if self.filter:
self.pipeline['mav_norm'] = Pipeline(
[windower, filter, fe, mmsc, nonneg])
else:
self.pipeline['mav_norm'] = Pipeline([windower, fe, mmsc, nonneg])
def polar():
num_channels = 5
dev = RandomWalkGenerator(rate=60,
num_channels=num_channels,
amplitude=0.03,
read_size=1)
# Polar plot can only show non-negative values
pipeline = Pipeline([Callable(lambda x: np.abs(x))])
Experiment(daq=dev, subject='test').run(
PolarPlotter(pipeline,
color=[0, 128, 255],
fill=True,
n_circles=10,
max_value=5.))
def bar():
num_channels = 10
channel_names = ['Ch ' + str(i) for i in range(1, num_channels + 1)]
dev = NoiseGenerator(rate=100,
num_channels=num_channels,
amplitude=5.0,
read_size=10)
pipeline = Pipeline(
[Windower(100),
Callable(lambda x: np.mean(x, axis=1, keepdims=True))])
Experiment(daq=dev, subject='test').run(
BarPlotter(pipeline=pipeline,
channel_names=channel_names,
group_colors=[[255, 204, 204]],
yrange=(-0.5, 0.5)))
def make_pipeline(self, rate, readsize, winsize, lowpass, highpass):
b, a = butter(4, (lowpass / rate / 2., highpass / rate / 2.),
'bandpass')
pipeline = Pipeline([
Windower(winsize),
# overlap = winsize - read_rate
Filter(b, a=a, overlap=winsize - readsize),
Callable(mean_absolute_value,
func_kwargs={
'axis': 1,
'keepdims': True
})
])
return pipeline