Single Source Dataset

The data set published here consists of simulated data time data of a microphone array simulating the flyby of a monopole or dipole source. The data was used in the European TRANSIT project (grant agreement number 881771) to test source separation methods.

The simulated measurement setup is shown in sketch.png in which the source passes the array at a distance of d=10m with a speed of 160 km/h. The source is simulated as a monopole or dipole source. Six different source waveforms are used which are indicated as Case 1.1 to Case 1.6 in the table. The amplitude of the source signals is rms = 1 at one metre from the source for all cases. The sampling frequency is 51200 Hz and a sound velocity of c=343 m/s was assumed.  The source starts its flyby at dx + d, where dx=8.68m is an additional distance, so that the source is at d after 10000 samples. All signals are 43040 samples long and simulate the flyby of the source over a time period of 1.19 s. The source moves over a total distance of 52.87 m. The microphone signals were recorded using a microphone with a frequency of 51200 Hz. The microphone signals were simulated using the Python package Acoular. 

Case 

Source setup 

Signal 

1.1 

one moving monopole point source 

white noise 

1.2 

one moving monopole point source 

pure tone signal with frequency 1000 Hz 

1.3 

one moving dipole point source in rail direction 

white noise 

1.4 

one moving dipole point source orthogonal to the rail 

white noise 

1.5 

one moving dipole point source in rail direction 

pure tone signal with frequency 1000 Hz 

1.6 

one moving dipole point source orthogonal to the rail 

pure tone signal with frequency 1000 Hz 

The array used is a bird array stretched in the X-axis with 80 microphones, which is shown in zugvogel_TRANSIT_80_2020.png. 

The dataset is provided as Single_source_dataset.h5. It contains the simulated time data of the sound pressure in the 80 microphones under the keys "Case 1.1" to "Case 1.6". The corresponding microphone geometry is provided as zugvogel_TRANSIT_80_2020.xml file. It contains the x,y,z positions of all 80 microphones which are indicated as point 0 to point 80 in the file. For example, to read the time signals from sample s1 = 1000 to s2 = 2000 of Case 1.1 from microphone m=5, the following Python code could be used:

import h5py

f = h5py.File('./Single_source_dataset.h5', 'r')
s1 = 1000
s2 = 2000
m = 5
f['case_1.6'][m,s1:s2]