فا   |   En
Login
مشاهده‌ مشخصات مقاله

Unseen Unit Generation for Unit Selection-based Concatenative Speech Synthesis

Authors
  • Majid Namnabat
  • M. Mehdi Homayounpour
Conference دوازدهمین کنفرانس بین‌المللی سالانه انجمن کامپیوتر ایران
Abstract The unit selection speech synthesis method is accounted of concatenative speech synthesis methods that it produces more natural quality speech than other approaches. So developing and improvement of this approach attracts great attraction to produce more natural output speech. Synthesis unit is one of important factors that have great effect on output quality of synthesizer systems. Using phones as synthesis unit can cause more natural output, however the discontinuity at the boundaries of concatenated phonemes can be potentially increased in output speech. In this paper, we proposed using unseen unit generation idea to eliminate discontinuities at phoneme boundaries when phone instances are used as synthesis units. In proposed method, existence discontinuities at phoneme boundaries in output speech are firstly found and then if it possible, one of adjacent phonemes of these boundaries are replaced by a suitable unseen unit. To synthesizing every unseen unit, left half phone of one existence instance of corpus are merged by right half phone of other existence instance. Perceptual Evaluations show quality output improvement due to decreasing discontinuities at boundaries.
قیمت
  • برای اعضای سایت : 100,000 Rial
  • برای دانشجویان عضو انجمن : 20,000 Rial
  • برای اعضای عادی انجمن : 40,000 Rial

خرید مقاله