• The detection mode : detect the BTE precisely.
• The tracking mode : use the detection informations to reduce the field to explore to find the eyes again.

• Face movement detection to isolate the face (can be coupled with a skin-color detection algorithm).
• CFF (Circle Frequency Filter) to extract the BTE point ( BTE = Between-The-Eyes). -- DEPRECATED
• A powerful local maxima extractor using grayscale images (fast hybrid grayscale resonctruction algorithm). -- DEPRECATED
• A connected components detection algorithm using bynary images.
• Using the algorithm to detect eyes with a incremental thresholding algorithm, and some heuristics about the eyes position in the face.

Return to Top

Screenshots

video_1.avi	Successful tracking of the eyes and the calculation of the "Between-The-Eyes" point.
video_2.avi video_3.avi	The processing of the BTE coordinates.
video_4.avi	The successful tracking and detection of head moves, nodding and head-shaking.
video_5.avi video_6.avi	The successful run of the application with mouse mapping system.

• Implement the skin-color detection algorithm using histograms.
• Finish the BTE movement interpretor.
• Take into account the blinking of the eyes.

• OpenCV
• Microsoft basic libraries (list later)

Change log

Version 0.6.1:

• Added a small window (radar) to see the position of the BTE compared to the center of mouse mapping.

Version 0.6.0:

• Added a new mouse mapping system, handling smooth moves and clics (simple right and left).

Version 0.5.6:

• Get rid of all the elements of OpenCV interface.
• New interface with separated configuration an display window.
• Possiblity of not displaying statistics and images for CPU savings.

Version 0.5.5:

• Operational head-shaking and nodding detection.
• Add a mouse position interpolation thread to have smooth mouse moves.

Version 0.5.4:

• First version of a direct move mouse mapper.

Version 0.5.3:

• Added state calculation to the interpretor.
• Fist version of head-shaking and nodding detection.

Version 0.5.2:

• First version of the move interpretor (direct interpretation).

Version 0.5.1:

• The incremental algorithm becomes pseudo-dichotomic.

Version 0.5.0:

• Use incremental algorithm using the connected component analysis to find the pupils.

Version 0.4.1:

• Add a statistics display.

Version 0.4.0:

• Not use CFF method anymore.
• Use of only Connected Components Analysis.

Version 0.3.5:

• Refining the heuristics for finding the eyes on the face.

Version 0.3.4:

• Connected components detection algorithm for grayscale pictures.

Version 0.3.3:

• BTE location calculator.
• Local maxima picture interpretor.

Version 0.3.2:

• Local Maxima extractor using a fast-hybrid grayscale reconstruction algorithm.

Version 0.3.1:

• CFF algorithm.

Version 0.3.0:

• OpenCV integration.
• Face dectection using simple algorithm (difference image analysis).

Version 0.2.2:

• The tracker is now an independent thread.
• FPS counting thread added.

Version 0.2.1:

• Successful end of the merge operation.

Version 0.2.0:

• Beginning of the merge of interface and tracker.

Version 0.1.0:

• First version of the application: interface and tracker are separated projects.
• First use of OpenCV library.
• First use of WxWidgets for the interface.

Whitepapers

• V. Graveleau, N. Mirenkov, G. Nikishkov, A Head-Controlled User Interface, Proc. of HC2005: The Eigth Human and Computer International Conference, The University of Aizu, Sept. 2005, p. 306-311.