def main_cropped_lms():
sns.set_style('darkgrid')
emg = load_emg_stack(
r'C:\Users\hbkm9\Documents\Projects\CYB\Balint\CYB104\Data',
task='Stand')
emg = [norm_emg(cur_emg) for cur_emg in emg]
targ = 5
cropped = [
crop_to_peaks(cur_emg[targ, :], cur_emg[7, :], 4, 0) for cur_emg in emg
]
return
def main_ecg():
emg = load_emg_stack(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment2\CYB102\Data',
task='Walk')
sns.set_style('darkgrid')
sns.set_context('poster')
lat = np.empty((0, 300))
ecg = np.empty((0, 300))
for f in emg:
pr = peak_regs(f[7, :], 100, 200, distance=800)
lat = np.vstack((lat, f[4, pr]))
ecg = np.vstack((ecg, f[7, pr]))
to_plot = norm_emg(np.vstack([np.mean(ecg, axis=0), np.mean(lat, axis=0)]))
fig: plt.Figure = plt.figure(constrained_layout=True)
spec = GridSpec(ncols=2, nrows=2, figure=fig)
ax = fig.add_subplot(spec[0, :])
ax.plot(np.linspace(-50, 99, 300), to_plot[0, :], ls='--')
ax.plot(np.linspace(-50, 99, 300), to_plot[1, :])
ax.get_yaxis().set_ticklabels([" "])
ax.set_ylabel("Arbitrary Units")
ax.set_xlabel("Time Relative to R peak (ms)")
plt.legend(["$\mathbb{E}\{V_{ECG}\}$", "$\mathbb{E}\{V_{Latissimus}\}$"])
ax2 = fig.add_subplot(spec[1, 0])
ax3 = fig.add_subplot(spec[1, 1])
def my_spec(ts_in, ax):
f, t, Sxx = sp.spectrogram(ts_in,
2000,
nperseg=256,
noverlap=128,
nfft=500)
ax.pcolormesh(t, f[:50], np.log10(np.square(Sxx[:50, :])))
ax.set_ylabel('Frequency [Hz]')
ax.set_xlabel('Time [sec]')
ax.set_ylim((0, 200))
my_spec(emg[20][5, :], ax2)
ax2.set_title('Latissimus Spectrogram')
my_spec(emg[20][7, :], ax3)
ax3.set_title('ECG Spectrogram')
peaks, _ = sp.find_peaks(emg[20][7, :],
distance=800,
height=np.max(emg[20][7, :]) * 0.5)
ax2.plot(peaks / 2000, np.ones_like(peaks) * 40, 'k+')
ax3.plot(peaks / 2000, np.ones_like(peaks) * 40, 'k+')
plt.show()
print('yee')
def main_spectrum_lms():
emg = load_emg(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment1\CYB005\Data\005_Walk20.json'
)
emg = norm_emg(emg)
step_dict = load_dict(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment1\CYB005\Data_p\005_ParamWalk20.json'
)
def consecutive(data, stepsize=1):
return np.split(data, np.where(np.diff(data) != stepsize)[0] + 1)
cl = np.array(step_dict["step_class"])
step_idx = consecutive(np.where(cl == 0)[0])
peaks, _ = sp.find_peaks(emg[7, :], distance=800, height=2.2)
targ = 6
plt.plot(emg[targ, :])
plt.figure()
n_bins = 50
F = spectrum_lms(emg[targ, :], n_bins, gamma=0.01)
#spec = norm_emg(np.abs(F.W[1:int(F.W.shape[0]/4), 3:]).T).T
#spec = spec-np.min(spec)+1
import seaborn as sns
sns.set_style('darkgrid')
sns.set_context('poster')
spec = np.log10(np.square(np.abs(F.W[1:int(F.W.shape[0] / 8), 1::40])))
plt.figure(figsize=(8, 5))
plt.pcolormesh(np.linspace(0, spec.shape[1] / 50, spec.shape[1]),
np.arange(spec.shape[0]) / n_bins * 2000, spec)
plt.xlabel('Time (s)')
plt.ylabel('Frequency (Hz)')
plt.tight_layout()
plt.vlines([st[0] / 100 for st in step_idx],
ymin=0,
ymax=(spec.shape[0] - 1) / n_bins * 2000,
label='Stride start')
plt.vlines([p / 2000 for p in peaks],
ymin=0,
ymax=(spec.shape[0] - 1) / n_bins * 2000,
lw=1,
color='w',
label='Heartbeat')
plt.legend()
plt.show()
return
def main_ar():
sns.set_style('darkgrid')
# emg = load_emg_stack(r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment1\CYB004\Data', task='Walk')
# targ = 0
# x0 = np.array([cur_x[targ, i*400:(i+1)*400] for cur_x in emg for i in range(int(cur_x.shape[1]/400))])
# from statsmodels.tsa.stattools import pacf
# pacs = [pacf(cur_seg, method='ywm') for cur_seg in x0]
# plt.stem(np.mean(pacs, axis=0), use_line_collection=True)
# plt.axhline(1.96 / np.sqrt(x0.shape[1]), color='red', ls='--')
# plt.axhline(-1.96 / np.sqrt(x0.shape[1]), color='red', ls='--')
emg = norm_emg(
load_emg(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment1\CYB004\Data\004_Walk12.json',
task='Walk'))
targ = 0
ord = 4
mu = 0.01
x = emg[targ, :]
X = ar_stack(x, ord)
F = LMS(x_in=X[:, :-1], d_in=x[ord:], mu=mu)
F.run()
plt.plot(F.W.T)
plt.show()
def main_anc():
sns.set_style('darkgrid')
sns.set_context('poster')
emg_stack = load_emg_stack(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment2\CYB102\Data',
task='Stair')
for i in range(4, 5):
emg = emg_stack[i]
emg = norm_emg(emg)
s = emg[4, :]
ref = emg[7, :]
filt_ord = 4
mu = 0.5
F = anc(s=s,
ref=ref,
beta=1,
order=filt_ord,
mu=mu,
eps=1 / mu,
pad=True,
rho=0,
act='tanh',
pretrain=(400, 3))
plt.figure()
plt.plot(np.linspace(0, len(s) / 2, len(s)), s * 4, lw=2)
plt.plot(np.linspace(0, len(s) / 2, len(s)), ref, lw=2)
plt.plot(np.linspace(0, len(s) / 2, len(s)), F.e * 4, lw=2)
plt.xlabel('Time (ms)')
plt.ylabel('Voltage (arbitrary units)')
plt.legend(('Trapezius signal', 'ECG reference', 'Filtered signal'))
plt.gca().get_yaxis().set_ticklabels([" "])
plt.tight_layout()
# plt.figure()
# plt.plot(F.W.T)
plt.show()
def main_spectrum_visu():
# region Loading and setup
sns.set_style('darkgrid')
sns.set_context('poster')
emg = load_emg(
r'C:\Users\hbkm9\Documents\Projects\CYB\Experiment1\CYB005\Data\005_Walk05.json',
task='')
emg = norm_emg(emg)
fig, axes = plt.subplots(2, 4)
names = [
'L Internal Oblique', 'R Internal Oblique', 'L External Oblique',
'R External Oblique', 'L Trapezius', 'R Trapezius', 'Erector Spinae',
'ECG'
]
for i in range(8):
if i is not 7:
axes.flatten()[i].plot(emg[7, :], alpha=0.6)
axes.flatten()[i].plot(emg[i, :])
if i is 0:
axes.flatten()[i].legend(['ECG', 'EMG'])
axes.flatten()[i].set_title(names[i])
plt.show()
ecg = emg[7, :]
emg1 = emg[5, :]
emg = emg[6, :]
# endregion
def plot_stuff():
# region Spectral plotting
fig, axs = plt.subplots(1, 3)
def my_spec(ts_in, ax):
f, t, Sxx = sp.spectrogram(ts_in,
2000,
nperseg=256,
noverlap=128,
nfft=2000)
ax.pcolormesh(t, f, np.log(Sxx))
ax.set_ylabel('Frequency [Hz]')
ax.set_xlabel('Time [sec]')
ax.set_ylim((0, 500))
my_spec(emg, axs[0])
axs[0].set_title('Trapezius Spectrum')
my_spec(emg1, axs[1])
axs[1].set_title('Erector Spinae Spectrum')
my_spec(ecg, axs[2])
axs[2].set_title('ECG Spectrum')
# endregion
# region ECG ARMA modelling
peaks, _ = sp.find_peaks(ecg, distance=800, height=2.2)
if min(peaks) < 59:
peaks = peaks[1:]
if max(peaks) > len(ecg) - 202:
peaks = peaks[:-1]
axs[2].plot(peaks / 2000, np.ones_like(peaks) * 100, 'k+')
axs[0].plot(peaks / 2000, np.ones_like(peaks) * 100, 'k+')
axs[1].plot(peaks / 2000, np.ones_like(peaks) * 100, 'k+')
plt.figure()
plt.plot(ecg)
qrs = np.zeros((len(peaks) - 1, 260))
for i in range(len(peaks) - 1):
qrs[i, :] = ecg[peaks[i] - 59:peaks[i] + 201]
plt.plot(peaks, np.ones_like(peaks) * 5, 'r+')
# endregion
from statsmodels.tsa.stattools import pacf
pacs = [pacf(cur_ecg, method='ywm') for cur_ecg in qrs]
plt.figure()
plt.stem(np.mean(pacs, axis=0), use_line_collection=True)
a = plt.axhline(1.96 / np.sqrt(qrs.shape[1]),
color='tab:red',
ls='--',
label='5% significance')
plt.axhline(-1.96 / np.sqrt(qrs.shape[1]), color='tab:red', ls='--')
plt.legend()
plt.tight_layout()
plt.xlabel('Delay (sample)')
plt.ylabel('Partial Autocorrelation')
plt.title('ECG Partial Autocorrelation Function')
plt.show()
plot_stuff()
emg1 = np.clip(emg1, -8, 8)
F = anc(emg1,
ecg,
5e-04,
rho=1e-08,
delay=0,
order=4,
mode='GASS',
alg='A&F',
alph=0.9)
plt.plot(emg1, alpha=0.1)
plt.plot(ecg / 4, alpha=0.1)
plt.plot(np.squeeze(F.W).T)
plt.show()
return