EyeTalk ( for more information see web presentation )

Our main goal is to build CNS in which our multimodal interfaces are embedded. The figure depicts our final product. The multimodal interfaces will be embedded in CNS. We will concentrate future work on enhancing the overall system speed, performance, and robustness. Optimal integration of multimode should be considered through many different approaches. Commercialization will come true if the system could be produced at low-cost and successfully embedded into CNS.

Final application of the experimental multimodal system in CNS

 

• Current Stage

  A sample output of the experimental PC-based multimodal system. Each component is executed in parallel.

• Current Performance

  Speech	Speed - 0.3sec/word	Word Recognition - 95%
  Lip	Speed - 10 frames/sec	Word Recognition - 50%(off-line test)
  Gaze	Speed - 5 frames/sec	Gaze Recognition - screen resolution 3 x 3
  Integration	Speed - 4~5frames/sec	Overall Command Recognition - 70 ~ 80%
  We have been trying to enhance overall recognition rate up to 95%.

กก

Back to Home