Distortion of BeamformingBase+ BeamfromerCleanSC

Ricardo

unread,

Dec 15, 2021, 4:40:47 AM12/15/21

to Acoular-users

Hi everybody,

I am working on a project that tries to compute DoA Maps of moving vehicles. We are using Acoular as the basis of our sound source localization.

In our algorithm we calculate the BeamformerBase plus the BeamformerCleanSC and plot the results over time so that you get a 2D DoA aggregated over a 10 sec time snippet.

If a vehicle passes by the sensor array (circular array) in a straight drive by, a problem arises: The results of the DoA for x-coordinates are reproduced properly, however y-coordinates are very much distorted in a fisheye-distorted manner.

Does anybody have an idea how this problem could arise?

Does anybody face the same problem when computing 2D DoA?

Thanks for your answers!

Gert Herold

unread,

Dec 15, 2021, 5:29:02 AM12/15/21

to Acoular-users

Could you provide an example image of what you're plotting? Using Clean-SC on a non-stationary source will probably lead to unsatisfactory results. With the BeamformerBase you always have to take into account that the map contains artefacts from the point spread function, which may add up in weird ways if the source is moving.

Ricardo

unread,

Dec 16, 2021, 4:33:06 AM12/16/21

to Acoular-users

Hi Gert,

thanks for your quick response! Unfortuneatly I can not share a detailed picture as I am most probably not allowed to do so by my company. Anyways I attached a draft of the problem.

The green trajectory is the ground truth of a vehicle driving by in lane 1 (roughly 4 m in front of sensors) and the red trajectory is the output of BeamformerCleanSC. The y-distance in the read trajectory is neither the real 4 meters nor is the x-distance in range of (-8m, 8m) as supposed to.

I hope that helps!

Distorted_traj_draft.png

Ennes Sarradj

unread,

Dec 16, 2021, 7:30:50 AM12/16/21

to Acoular-users

I am afraid I really do understand what you are trying to do. If the source is moving, it is a very bad idea to use frequency domain beamformers. I would not expect meaningful results, especially not from CleanSC. You can use time-domain beamforming and if you want not just DOA but also range you should experiment with the different steering vector formulations.

Ricardo

unread,

Dec 17, 2021, 2:35:12 AM12/17/21

to Acoular-users

Hi all,

thanks for the quick responses, I will try to adapt my algorithm to a time domain beamformer and I will try to use different steering vector fomrulations. Once I have meaningful results I willl come back to you.

I am sorry about not beaing able to share more information, I will have to double check with my company If I am allowed to share more details!