Machine Lipreading of Concatenated Vowels Based on a Coarticulation Model.

DOI Web Site Web Site

Bibliographic Information

Other Title
  • 調音変換モデルに基づく連母音口形識別
  • チョウオン ヘンカン モデル ニ モトズク レンボイン クチガタ シキベツ

Search this article

Description

This paper discusses characteristics of coarticulation in terms of lip shape and proposes a coarticulation model for lipreading coarticulated vowels. The characteristics are described by the width and height of the lip shape and the distance between the upper lip and the bottom of the jaw. The model consists of a conversion matrix and an average shift vector for the above three parameters. The conversion matrix converts the covariance matrix of parameters for an isolated vowel to that for its coarticulated vowel. The average shift vector shifts the average vector for an isolated vowel to that for its coarticulated vowel. The effectiveness of the model is demonstrated.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top