santi-pdp/segan

Questions about voice pesq

xiaojian10 opened this issue · 0 comments

Hello, this is a great article. I downloaded your weight file and tested some noise files, but I found that the processed voice file has a low pesq, but it sounds good. The three maps below use this as clean_speech, noise_speech, processed_speech.
image
image
image
I don't know why the speech index after processing is very low. Can you give me some things to pay attention to when testing the index?Or can you share your code for testing the PESQ indicator?