[Eilish N]

Hi, I have just come across this post!

I’m looking at fixing pitch marking errors for my first investigation – in particular errors that arise because of an erroneous prediction in the utts.mlf file of a “sil” phone mid-utterance (which should just be adequately handled by sp). I have managed to get rid of ~300 pitchmarking errors just by deleting these inter-utterance sil phones from the utts.mlf file and re-running the forced alignment – so supposedly the alignment of phones to the database is now much better.

I want to test if this actually has any effect on the output voice (I suspect that maybe the effect will be too small to be noticed, especially on in-domain sentences where contiguous units are often chosen anyway). Is the best way to go about this listening test by building two voices made up of only the originally bad pitchmarked utterances – one with pitchmarks fixed and one not, and then choosing test sentences that should highlight the error?

[Eilish N]

Hi Catherine,

Thank you so much that’s working now! And has solved the sound issue as well 🙂 Thank you!

[Eilish N]

Hi Catherine,
This is the output I got:
ls:cannot access ‘/Volumes/Network/courses/sp/assignment1/’: No such file or directory

I’ve attached the screenshot as well!

Attachments:

You must be logged in to view attached files.

[Eilish N]

Thank you! I’ve contacted them. I’m now having trouble with this command:
cp /Volumes/Network/courses/sp/assignment1/config.scm .
On the remote desktop it is saying that no such file or directory exists. Is this because I don’t have access?

[Author]

Posts