BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:speech synthesis seminar series
SUMMARY:Statistical Parametric Speech Synthesis Based on S
 peaker and Language Factorization - Heiga Zen (Tos
 hiba Research Europe Ltd.)
DTSTART;TZID=Europe/London:20110621T130000
DTEND;TZID=Europe/London:20110621T143000
UID:TALK31796AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/31796
DESCRIPTION:An increasingly common scenario in building hidden
  Markov model-based speech synthesis and recogniti
 on systems is training on inhomogeneous data.  For
  example\, data from multiple different sources an
 d/or different types of data are used.  This semin
 ar introduces a new technique for training hidden 
 Markov models on such inhomogeneous speech data\, 
 in this case including speaker and language variat
 ions. The proposed technique\, speaker and languag
 e factorization\, attempts to factorize speaker-sp
 ecific/language-specific characteristics in the da
 ta and model them by individual transforms.  Langu
 age-specific factors in the data are represented b
 y transforms based on cluster mean interpolation w
 ith cluster-dependent decision trees.  Acoustic va
 riations caused by speaker characteristics are han
 dled by transforms based on constrained maximum li
 kelihood linear regression.  This technique allows
  multi-speaker/multi-language adaptive training to
  be performed.  Since each factor is represented b
 y an individual transform\, it is possible to fact
 or-in only one of them.  Experimental results on s
 tatistical parametric speech synthesis show that t
 he proposed technique enables the speaker and lang
 uage to be factorized\, allowing the speaker trans
 form estimated in one language to be successfully 
 used to synthesize speech in different language wh
 ile keeping the voice characteristics.\n
LOCATION:Cambridge University Engineering Department\, Lect
 ure Room 11
CONTACT:Kai Yu
END:VEVENT
END:VCALENDAR