Proceesing received audio signals for MUSIC DOA

[nonster666]

I am trying to implement MUSIC DOA estimator for audio signals.
I all papers, forums and code examples I have seen (including here https://www.edaboard.com/showthread.php?t=138861), the received signals is multiplied by an array manifold derived from the theoretical direction from which the signals was received and then noise is added (i.e., the received signal model X=S*A+N).
Thereafter the MUSIC algorithm is ran on X and the MUSIC spectrum is generated and plotted.
However, if I do not now the theoretical direction from which the audio signals were received (i.e., a real life situation), and I want to find the directions of two or more audio sources, what do I use for A (if at all) before running the music algorithm?
Thank you

 

[weetabixharry]

With MUSIC, you must search over "all" values of all parameters of interest. (Of course, you can't search all the infinitely many directions, so we instead evaluate a limited number of fixed directions -- for example 0°, 0.5°, 1°, 1.5°, ..., 359.5°).

I wrote some simple code with lots of comments at this post which should help to clarify how it works.

 

[nonster666]

Thank your weetabixharry, However I need some clarifications

The code you written

% Received signal
x = A*m + n;

i.e., m (the original signal) isbmultiplied by A and noise is added, where

A = exp(-1j*r*k)

k = pi*[cosd(az).*cosd(el), sind(az).*cosd(el), sind(el)].';

and
az = [35;39;127].

As I have written in my original post, this assumes that we know 'az' to generate the received signal.
However, in a real life situation we do not know 'az'.
what then do I use for 'x'?