[hts-users:03012] buiding pentaphone contextual features
- Subject: [hts-users:03012] buiding pentaphone contextual features
- From: Guillaume Galou <guillaume.galou@xxxxxxxxx>
- Date: Fri, 19 Aug 2011 16:00:04 +0200
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=subject:from:to:in-reply-to:references:content-type:date:message-id :mime-version:x-mailer:content-transfer-encoding; bh=uT0aSewTCPtcA40a5CZp8oSBRn/tNJ8ISJkoxfWpsRY=; b=PFIMVG+RzZ5eI4er7hws+ykAPs3a6t+ZXlAsRjJKfGlhz5AlKKB2C8zZtzm0LbxVs0 yBRuu0OEaIF2XK6sknSmcrXqEY6lbWcZwnzgI8VEC5h978lJnPHzEk9JHll+DC/diKmd rxhOJk/fWkP6Sy5M4OdenfvEDmk66MYFomoVc=
Dear all,
I'm trying to build a script in order to create contextual features from
raw texts inputs. Of course, I would like to do it using pentaphones
(2f2b) as shown in the demos.
To do this, I tried to:
- force align word-level transcription with the -m option of HVite -->
this gives aligned phones OK
- use HLEd in order to create associate tri-phone level using the TC
script command --> this gives triphone-based aligned transcription OK
It seems impossible to run HLEd in order get pentaphones directly (no
command like in the triphone case). I know this is not directly linked
to HTS but HTK. Nevertheless pentaphones are not that used in the speech
recognition area. I'm sure many of HTS users faced this problem.
The answer might be writing a script, I'm just trying to save time.
Best regards.
- References
-
- [hts-users:03007] Checking synthesized speech in demos, Huda Sarfraz