Example

[golamsadeghnia]

Hi there, I was wondering how to set up the sensor and source position parameters? Can you please give an example? Thank you!

[atsiami]

Hi,

It is written in the beginning of multichannel_speech_enhancement function:
Inputs:

.

.
#X: wavs, X(:,n) is the nth sensor's wav

sensor_positions: the positions of the sensors. sensor_positions(:, n)

are the coordinates of the n-th sensor.

source_position: the source position. source_position(:, 1) are the

source coordinates.

.
.
.
For example if you have three sensors positioned in (1, 0.5 ,1.70),(1.3,0.5,1.7),(1.6,0.5,1.7) and a source in (2,1,2) you would give the input as: sensor_positions=numpy.array([[1, 1.3, 1.6],[0.5, 0.5 0.5],[1.7,1.7,1.7]]) and source_position=numpy.array([[2],[1],[2]]).

Hope I helped!
Antigoni

[bghani]

Hi,

First of all thanks a lot for making the code public. I actually wanted to implement the MVDR beamformer and then I was glad to find your code. I am facing some difficulties understanding the noise field coherence matrix calculation part of your code. More specifically, I am unable to understand why you are reestimating the coherence matrix (second function) and then in the third function what limit are you putting it though. So that I know what should I chose my value of limit for the problem I am working on. I searched a lot and couldn't find any literature on how to calculate the noise field coherence matrix. Could you be kind enough and either explain this a bit or refer me to a paper or some reference where I can find how to compute the noise coherence matrix?

Thanks a lot.

Burooj

[Ding777]

Hi thank you for your code, I think the output of g_arr for the proposed method and others will be complex number(except for Mc cownan and Zelinski), since you used the mvdr weights which is complex, so did you take real values of g_arr to multiply with the signal input to get the output?