search

ppwwyyxx / speaker-recognition

A Speaker Recognition System
Apache License 2.0
675 stars 276 forks source link

Features scores issue #84

Open Schalkn opened 5 years ago

Schalkn commented 5 years ago

Hi I took over a project based on your libraries. I'm trying to determine whether two voice clips come from the same speaker or not. There's one thing I don't understand: At one place, all the features of the two voice clips are compared, and I get somewhere between 3000 and 7000 scores as output, ranging roughly between -120 and 100. How do I use these to determine how good/bad the match was?

ppwwyyxx commented 5 years ago

From the absolute score it's hard to asses a match. The scores are typically used for comparison with each other.

Schalkn commented 5 years ago

Thank you for your quick reply!! Really appreciate it. So how do I decide how good/bad they are matching? I also get some result called "average quality", which is eg -9, but I don't know if this is usable, or what it means?

Schalkn commented 5 years ago

Hi there. I'm so sorry to bother you so much (hope I don't bother you in your holiday as well). Can you perhaps help me, in how I can decide if two voice clips are a match, by using the scores? Or is it not possible at all? Thank you so much.

ppwwyyxx commented 5 years ago

I don't know if it's possible to decide a match from absolute score.

Schalkn commented 5 years ago

How do you do it?

From: Yuxin Wu Sent: Wednesday, 02 January 2019 22:07 To: ppwwyyxx/speaker-recognition Cc: Schalkn; Author Subject: Re: [ppwwyyxx/speaker-recognition] Features scores issue (#84)

I don't know if it's possible to decide a match from absolute score. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

Schalkn commented 5 years ago

Can you perhaps help? How do you do it?

ppwwyyxx commented 5 years ago

I don't know how to decide a match from absolute score.

Schalkn commented 5 years ago

So if getting the absolute scores are the wrong way, how should I compare two voice clips, to see if they are the same person?