Wrong Supported samplerate for Task 3: Target speaker extraction (16kHz)

[yezhangyinge]

I try "Sub-Task 1: Audio-only Speaker Extraction Conditioned on a Reference Speech", but found that

1. the supported samplerate is 8k and only work for 8k. If I input 16k audio, the model will output wrong result.
2. the model is really bad at separating 2 spks. Is the ckpt right here?

[zexupan]

Sorry we made a mistake, the "Sub-Task 1: Audio-only Speaker Extraction Conditioned on a Reference Speech" is 8kHz instead.

[zexupan]

For the performance, the model for this sub-task isn't trained on a large amount of data. It is only trained on WSJ0-2mix, so it may have a bad generalisation performance.

[yezhangyinge]

OK, I get it. Thanks for your reply! I would like to know if there will be a model that is trained on large amounts of data. Or will there be a model (except SpEx+ like BSRNN, TFGridNet) that has a better performance?