Tongue Segmentation with Geometrically Constrained Snake Model

Zhihua Su, Jianguo Wei, Qiang Fang, Jianrong Wang, Kiyoshi Honda


Articulatory visualization aims at providing precise visual information of the speech organs (tongue, lips and velum) that accompany with speech signals. It is often critical in fundamental studies and certain applications. To construct an articulatory visualization system, the profile of the speech organs must be segmented from images acquired by various types of medical equipments. In this paper, a geometrically constrained snake model is proposed to segment tongue profiles from mid-sagittal MRI to deal with the situation in which the tongue contacts with the surrounding structures and the target object with inhomogeneity nature. The result indicates that the proposed method improves segmentation performance significantly compared with the traditional snake model.


 DOI: 10.21437/Interspeech.2018-1108

Cite as: Su, Z., Wei, J., Fang, Q., Wang, J., Honda, K. (2018) Tongue Segmentation with Geometrically Constrained Snake Model. Proc. Interspeech 2018, 3117-3121, DOI: 10.21437/Interspeech.2018-1108.


@inproceedings{Su2018,
  author={Zhihua Su and Jianguo Wei and Qiang Fang and Jianrong Wang and Kiyoshi Honda},
  title={Tongue Segmentation with Geometrically Constrained Snake Model},
  year=2018,
  booktitle={Proc. Interspeech 2018},
  pages={3117--3121},
  doi={10.21437/Interspeech.2018-1108},
  url={http://dx.doi.org/10.21437/Interspeech.2018-1108}
}