akojimaslp / beamforming-for-speech-enhancement Goto Github PK

Thasnks for sharing your nice code, and I have a question about the number of frames in calculating correlation_matrix of MVDR, why not using all frames but just first 10 and last 10 frames of the audio sample?

Hi author, thank you for your great work, but I have some questions about mic_steering_vector.
Do you use a circular array？I can not understand the meaning of [0,60,120,180,270,330]. I look forward to your reply，thanks.

Hi,
Thank you for your nice code.

I have some questions. Please answer me.

1. How do you get the data?
It seems to be a .wav file of the CHiME dataset.

2. How do you get the MIC_ANGLE_VECTOR and LOOK_DIRECTION?
As I understand, MIC_ANGLE_VECTOR is assigned for each microphone.
And as I know, chime's channel 1~6 represent the mic (http://spandh.dcs.shef.ac.uk/chime_challenge/chime2015/overview.html).

But your MIC_ANGLE_VECTOR is hard to understand.
MIC_ANGLE_VECTOR = np.array([0, 60, 120, 180, 270, 330])

I'd liked to know the definitions and equations of the MIC_ANGLE_VECTOR and LOOK_DIRECTION.

3. Could I know your reference for the steering vector?
These two equations are the steering vector equation which I know.

h=exp(-j2\pi * freq / c * r * cos(theta)) [1]
h_i=exp(-j2\pi * freq / c* (r_i-r0)) / N * (r_i/r_0) [2]

As I understand, the first equation[1] assumes that the target source arrives at the microphone in the planar wave. And the second equation[2] assumes that the target source arrives at the microphone in the non-planar wave.

Especially, there is no usage of MIC_DIAMETER.

[1] https://courses.engr.illinois.edu/cs498ps3/sp2018/ewExternalFiles/Lecture%204%20-%20Microphone%20Arrays.pdf
(Page 18)
[2] Ennes Sarradj, "Three-Dimensional Acoustic Source Mapping with Different Beamforming Steering Vector Formulations", Advances in Acoustics and Vibration, vol. 2012, Article ID 292695, 12 pages, 2012. https://doi.org/10.1155/2012/292695
(Equation 10)

Thank you again for your nice code.

When I run this program, I found that the soundfile is missed.