regarding source seperation about pyroomacoustics

[fakufaku]

Hi, this is a pretty broad question. The quality of the separation will depend a lot on what you want to do and the constraints you are working with. For example, if the microphones are placed closely together, the separation performance will not be so good in the low frequency part.
One way to separate is to create beamformers towards the sources. Beamformers work by combining the microphone signals in a way that creates constructive interferences towards the sources. Now there are, as you might have noticed, many ways to do that.

[pathasreedhar]

Yes thats true, I experienced this as I am working with PS3EYE. Right now I am using webrtc beamforming module. I wanted to know how to use the beamformer module present in this pyroroomacoustics.

[pathasreedhar]

Hi,
I want to use beamforming code present in this repo, could you give me a glimpse of how to use it. I have matrix creator a circular array with 8 mics. I want to beamform and do source separation

[pathasreedhar]

Hi any help possible

[fakufaku]

Hi @pathasreedhar , there are two beamforming examples in the examples folder. Did you take a look there ?

[pathasreedhar]

yes.. I have seen them, these examples use simulations, I am unable to use them for real time readings, like for example in time domain, you used some beam forming filters suitable for simulated readings.
I want to know how we select these parameters

[fakufaku]

You can use the Beamformer object on recorded signal as well.

'''
This example shows how to create delay and sum beamformers
'''
from __future__ import print_function, division

import numpy as np
import matplotlib.pyplot as plt
from scipy.io import wavfile
import pyroomacoustics as pra

# Open your recorded signals
fs, my_recording = wavfile.read('my_recording.wav')

# Create a room around your array
# this is merely a convenience, it doesn't have to be the real room
room = pra.ShoeBox([4,6], fs=fs)

# Add a source somewhere in the room
# You don't need to know the exact location, just set
room.add_source(<coordinate of the source relative to your microphone array>)

# Create a linear array beamformer with 4 microphones
# with angle 0 degrees and inter mic distance 10 cm
R = <coordinates of the microphones on your array>
room.add_microphone_array(pra.Beamformer(R, room.fs))

# Add the recorded signals in your array
room.mic_array.signals = my_recording

# Now compute the delay and sum weights for the beamformer
# Use the far field option if you don't know the exact distance
# from source to microphone array
room.mic_array.rake_delay_and_sum_weights(room.sources[0][:1], far_field=True)

# alternatively you can compute your own weights and set them directly
# room.mic_array.weights = my_weights

# Now compute the output of the beamformer
output = room.mic_array.process()

I agree that the interface for the beamforming is not ideal at the moment. Hopefully, I can improve this in the future.

[fakufaku]

Closing this due to lack of activity.