MACHINE VISION RECOGNITION OF EMBLEMATIC FACIAL EXPRESSIONS TO FACILITATE HUMAN-COMPUTER INTERACTION

QUADRIPLEGICS AND OTHER SEVERELY DISABLED INDIVIDUALS WITH DIMINISHED MOBILITY BELOW THE NECK ARE LIMITED IN THE MANNER BY WHICH THEY MAY UTILIZE COMPUTERS AND INSTRUMENTATION THAT REQUIRE MANUAL MANIPULATION AND ARE THUS PRECLUDED FROM PURSUING A HOST OF EDUCATIONAL AND EMPLOYMENT OPPORTUNITIES. SINCE FACIAL EXPRESSIONS ARE GENERALLY EASY TO DISPLAY AND TO CONTROL, MAY MANIFEST OVER 100 DISTINCT CONFIGURATIONS, AND COMPRISE THE MAJORITY OF NONVERBAL CUES IN HUMAN INTERACTION, THE EXHIBITION OF FACIAL EMBLEMS (FACIAL EXPRESSIONS THAT REPLACE SPOKEN MESSAGES) IS A MORE VERSATILE, LESS FATIGUING, AND LESS ENVIRONMENTALLY INSTRUSIVE ALTERNATIVE FOR COMPUTER INPUT SPECIFICATION THAN THE TECHNIQUES REQUIRED BY EXISTING HANDICAP AIDS. TO EXPLOIT THIS MEDIUM TO FACILITATE HUMAN-COMPUTER COMMUNICATION BY THE DISABLED, AMERICAN RESEARCH CORPORATION OF VIRGINIA PROPOSES THE DEVELOPMENT OF A MACHINE VISION INTERFACE FOR TH CAPTURE AND INTERPRETATION OF USER-DEFINED FACIAL EMBLEMS. TECHNICAL OBJECTIVES INCLUDE THE SPECIFICATION AND SAMPLING OF EMBLEMATIC FACIAL EXPRESSIONS, THE DEVELOPMENT OF A FACE LOCATOR WHICH IS INVARIANT TO CHANGES IN BACKGROUND CONFIGURATION, THE DEVELOPMENT OF A FEATURE IDENTIFIER FOR SEGMENTATION OF THE PRIMARY NONVERBAL COMMUNICATION CHANNELS WITHIN THE FACE, THE EXTRACTION OF GEOMETRICAL ATTRIBUTES FROM SEGMENTED FACIAL FEATURES FOR NEURAL NETWORK ANALYSIS, AND FORMATIVE EVALUATION OF SYSTEM ACCURACY.