number of mics and audio samples

[nguyenvulong]

Hello,
Could you please explain a bit about the configuration.
I wonder if n_mic_keep and n_audio_keep should be identical?

For example

n_mic_keep = [5, 6]

means 5 or 6 mics will be used

in that case, a corresponding number of audio samples (either 5 or 6) should be used too, because each mic emits one signal (audio).

and if so, should the code have some constraint about these two values?
(for example, checking whether n_mic_keep and n_audio_keep are the same)

[axeber01]

n_mic_keep = [mic_min, mic_max] and n_audio_keep = [audio_min, audio_max] can actually be different. They control the ranges for how many coordinates and audio snippets will be masked during training. The following scenarios can happen for each microphone:

1. both coordinate and audio is kept as input
2. only coordinate is kept as input, audio is masked
3. only audio is kept as input, coordinate is masked

however, we never mask both coordinate and audio for the same microphone, as this would lead to redundancy.

there are not any formal restrictions on n_mic_keep and n_audio_keep , but we need at least 4 or 5 microphones with both audio and coordinates to make localization tractable. Therefore it is reasonable to use something like min(mic_min, audio_min) = 5.

[nguyenvulong]

I see. That's a good take.
But can you please explain why this would lead to redundency?
What kind of "redundency" is it?

however, we never mask both coordinate and audio for the same microphone, as this would lead to redundancy.