Sketch misrecognition correction system based on eye gaze monitoring

A gaze based error recognition system comprising:

• at least one computer, an interaction surface, an eye tracker to capture natural eye gaze behavior during sketch based interaction on said interaction surface, and a sketch-based interface configured to provide a recognition feedback to interpret user drawn sketches on said interaction surface, wherein;
• said sketch based interface comprises a prediction module;
• said prediction module comprises a multimodal data collection module configured to collect real-time synchronized user gaze data, captured by said eye tracker and sketch data in relation to said interaction surface and to segment said collected gaze data into fixations and saccades;
• said prediction module comprising a feature extraction module configured to determine, from said fixations of the segmented gaze data, characteristic features of user eye movement in a form of duration of last fixation to the recognition feedback, duration of last visit to the recognition feedback, mean duration of fixations and mean duration of visits;
• said prediction module comprises a two-class classifier configured to effect classification of the gaze data into correction intending and general purpose fixation classes according to the values of their characteristic features;
• and
• to provide a prediction result as to whether said recognition feedback is correct or not based on the class assigned by said classifier, and
• wherein said classifier consists of a machine learning algorithm and training data for said machine learning algorithm consists of the gaze data and sketch data of users recorded during sketch-based interaction.