def load_channels(dirs):
"""
Load all the 868 channels
If it is a continious measurement we split the channel snapshots per 100 snapshots
All locations are stored in an array and returned per antenna conf ULA/URA
"""
H_ula = []
H_ura = []
for d in dirs:
# only process 868 measurements
is_868_b = is_868(d)
is_ula_b = is_ula(d)
if not is_868_b:
print(f"Skipped {d}")
continue
print(f"processing {d}")
H = np.load(os.path.join(root_dir, d, "small-channel.npy"))
# remove faulty antenna 32
H = H[:, :, :-1]
H = normalize(H)
if is_cont_meas(d):
# too much samples, split in 1 sec snapshots, i.e. 100 snapshots
H_splitted = np.array_split(H, H.shape[0] // 100)
for H_split in H_splitted:
if is_ula_b:
H_ula.append(H_split)
else:
H_ura.append(H_split)
else:
if is_ula_b:
H_ula.append(H)
else:
H_ura.append(H)
return H_ula, H_ura
import numpy as np
from tqdm import tqdm
from utils.load_yaml import load_root_dir
from utils.util_loc import is_cont_meas, is_ula, is_868, get_path, get_point
import matplotlib.pyplot as plt
if __name__ == '__main__':
root_dir = load_root_dir()
subdir, dirs, files = next(os.walk(os.path.join(root_dir)))
ax = plt.gca()
res = []
for d in dirs:
# only process 868 measurements
is_868_b = is_868(d)
is_ula_b = is_ula(d)
if not is_868_b:
# print(f"Skipped {d}")
continue
# print(f"processing {d}")
H = np.load(os.path.join(root_dir, d, "small-channel.npy"))
# remove faulty antenna 32
# [snapshots x freq points x BS antennas]
H = H[:, 0, :31]
if (not is_cont_meas(d)) and (is_ula(d)):
H = np.nanmean(H, axis=0)
# H = np.mean(H, axis=0)
H_norm = H / np.linalg.norm(H)
import random
import numpy as np
from utils.load_yaml import load_root_dir
from utils.util_loc import is_cont_meas, is_ula, is_868
import matplotlib.pyplot as plt
import aoa_algorithms as alg
root_dir = load_root_dir()
subdir, dirs, files = next(os.walk(os.path.join(root_dir)))
if __name__ == '__main__':
for d in dirs:
if is_868(d) and not is_cont_meas(d) and is_ula(d):
H = np.load(os.path.join(root_dir, d, "small-channel.npy"))
H = H[:, :, 0:31]
# average over freq points
H_avg = np.mean(H, axis=1)
H_avg = H_avg.transpose() # so it is a MxN matrix
cov_mat = H_avg @ H_avg.conj().transpose()
M = H.shape[2]
angles = np.linspace(-np.pi / 2, np.pi / 2, 360)
pspectrum, psindB, peaks = alg.music(cov_mat, 1, M, angles)
plt.cla()
plt.plot(angles, psindB)
plt.plot(angles[peaks], psindB[peaks], 'x')
plt.legend(['pseudo spectrum', 'Estimated DoAs'])
plt.savefig(os.path.join(root_dir, d, "power_spectrum.png"))