Context
The context of this work is in articulatory tracking, with applications in speech therapy, linguistics, and human-computer interaction.
Results
We introduced an innovative approach to tongue tracking in ultrasound images by integrating a biomechanical model. Unlike purely image-based methods, we used a finite element model to simulate the physical behavior of tongue tissue. The method incorporates both surface points visible in the ultrasound and internal anatomical landmarks. They are tracked by an optical flow method, with each observation assigned a confidence level based on tracking quality. These uncertainties are used in the biomechanical model as soft constraints, allowing the tongue deformation to be guided by reliable data while remaining robust to noise [1].
|
 |
