MUSIC DOA Estimation for a circular array with real time audio signals

Hi,

Iam working on an application of speech signals DOA estimation. I started my implementation with MUSIC algorithm. My circular array is a uniform array of 6 element with 60 degree spacing for each mic. I wanted to estimate the angle of arrival of speech signals when talking in front of a circular array. So Iam not using any signal collector functions like collector, Widebandcollector that are available in matlab instead I am using real time speech signals from microphones and loading it to the algorithm. But everytime Iam getting an estimation of zero degree.

From many approaches I found to prepare input data for estimation take the FFT of the signal and then proceed with covariance matrix calculation but I am little confused here. How to create the covariance matrix from the FFT values? If I take the FFT complex values from each mic for the same frequency I would be able to create a num_mics x num_mics covariance matrix. But which frequency bin I have to select to do the same?

I am getting an estimation of zero degree every time.

Will someone suggest the steps clearly so that I can proceed, or did I miss any steps here?

Please suggest any solution for the same.

[David Meissner]

If it was me, I would try implementing the example here (https://www.mathworks.com/help/audio/ug/live-direction-of-arrival-estimation-with-a-linear-microphone-array.html) with simulated data, then with your audio stream from a 1D array, and then try converting it from a 1D array (as in the example) to a 2D array (your circular).