In [81]:
import numpy as np
import matplotlib.pyplot as plt
from scipy.io import wavfile
import json
plt.rcParams['figure.figsize'] = 15, 5

Data for Direction of Arrival with MUSIC

A part of Statistical Signal and Data Processing through Applications class at EPFL

Experiment description

The experiment details are described at https://github.com/fakufaku/pyramic-dataset/blob/master/PROTOCOL.md. The geometry of the experiment is described in the machine readable file https://github.com/fakufaku/pyramic-dataset/blob/master/protocol.json. Below is an example how to get the indexes of the top layer of microphone for 2D experiments.

In [87]:
with open('protocol.json') as json_file:  
    protocol = json.load(json_file)
    
microphone_3D_locations = np.array(protocol['geometry']['microphones']['locations'])
top_mics = np.isclose(microphone_3D_locations[:,2], 0.06123724)
print("Total {} microphones on top flat layer,".format(np.sum(top_mics)))
print("at indexes indexes: ", np.arange(len(top_mics))[top_mics])

microphone_2D_locations = microphone_3D_locations[top_mics, :2]

speaker_distance = protocol['geometry']['speakers']['location']['middle']['distance']
print("First speaker distance from the center of the array: {}m".format(speaker_distance))

Total 24 microphones on top flat layer,
at indexes indexes:  [ 8  9 10 11 12 13 14 15 24 25 26 27 28 29 30 31 40 41 42 43 44 45 46 47]
First speaker distance from the center of the array: 3.944m

Data

Instructions how to download and use data are at https://github.com/fakufaku/pyramic-dataset. We are interested in all segmented recordings.

Since we are interested in the 2D case, we are interested only in the first speaker (index 0), so you can remove files that have spkr1 and spkr2 in the name.

For each speaker and type of sound, there are 180 different directions of arrival available to test the algorithm. Since each file contains sound from only one mike, in order to test algorithm on multiple sources, you have to add recordings from different files.

Below is an example how to load a single .wav file. The file contains recordings of all 48 microphones, and you need only the top ones. You can also take only some of the top micrphones, but remember that they can't all lay one one line.

In [88]:
rate, data = wavfile.read('segmented/sweep_lin/sweep_lin_spkr0_angle262.wav')
print("sampling rate: {}".format(rate))
data_2D = data[:, top_mics]
plt.plot(data_2D[:400,:4])
plt.title("Beginning of the recording")
plt.show()

sampling rate: 48000
In [89]:
plt.plot(data_2D, alpha = 0.5)
plt.title("All 2D data")
plt.show()

You dont need to test data on all angels, types of signals and configurations. Pick a few and justify your choices. Have fun! :)

In [ ]: