I'm using doatools.py library (https://github.com/morriswmz/doatools.py) Now, my code looks like:
import numpy as np
from scipy import constants as const
import math
import doatools.model as model
import doatools.estimation as estimation
def calculate_wavelength(frequency):
return const.speed_of_light / frequency
# Uniform circular array
# X
# |
# X---------X
# |
# X
NUMBER_OF_ELEMENTS = 4 # elements are shown as "X"
RADIUS = 0.47 / 2
FREQ_MHZ = 315
freq = FREQ_MHZ * const.mega
wavelength = calculate_wavelength(freq)
antenna_array = model.UniformCircularArray(NUMBER_OF_ELEMENTS, RADIUS)
# Create a MUSIC-based estimator.
grid = estimation.FarField1DSearchGrid()
estimator = estimation.MUSIC(antenna_array, wavelength, grid)
R = np.array([[1.5, 2, 3, 4], [4, 5, 6, 5], [45, 5, 5, 6], [5, 1, 0, 5]])
_, estimates = estimator.estimate(R, 1, return_spectrum=False, refine_estimates=True)
print('Estimates: {0}'.format(estimates.locations))
I can generate signal with this library, but how to use my own? For example, signal from ADC (like this:
-> Switching to antenna 0 : [0, 4, 7, 10]
-> Switching to antenna 1 : [5, 6, 11, 83]
-> Switching to antenna 2 : [0, 23, 2, 34]
-> Switching to antenna 3 : [23, 105, 98, 200]
)
I think your question is how you should feed the real data from antennas, right? Supposedly your data should be in order along time. I mean in case of "antenna 0 : [0, 4, 7, 10]", 0 is the 1st-in data, and 4, 7, in order, and the 10 is the last one in time. If yes, you could leave them as a simple matrix like what you typed above:
r = matrix 4x4 of
0, 4, 7, 10
5, 6, 11, 83
0, 23, 2, 34
23, 105, 98, 200
//===============
r(0,0) = 0, r(0,1) = 4, r(0,2) = 7, r(0,3) = 10
r(1,0) = 5, r(1,1) = 6, ... etc.
r(2,0) = 0, ...etc.
//==============
R = the product of r and its hermitian matrix (r.h in python).
R = r @ r.h
And this is the covariance matrix that you need to fill in as the 1st argument in function.