VOICE I/O FOR FLIGHT MANAGEMENT SYSTEMS

THIS EFFORT IS INTENDED TO DEMONSTRATE THE FEASIBILITY OF DESIGNING A VOICE I/O (INPUT/OUTPUT) SYSTEM SUITABLE FOR APPLICATION IN THE COCKPIT OF A COMMERCIAL AIRCRAFT FOR THE PURPOSE OF FACILITATING INTERACTION WITH A SOPHISTICATED FLIGHT MANAGEMENT SYSTEM. THE ULTIMATE SYSTEM NEEDS TO INCORPORATE (1) A SPEAKER-INDEPENDENT, NOISE-RESISTANT, REAL-TIME, CONTINUOUS SPEECH RECOGNIZER WITH NEAR NATURAL LANGUAGE CAPABILITIES, PREFERABLY ONE THATDOES NOT REQUIRE SPECIAL PURPOSE HARDWARE OR EXCESSIVE COMPUTING POWER; AND (2) A SPEECH SYNTHESIZER THAT IS BOTH EASILY UNDERSTOOD AND CLEARLY IDENTIFIABLE AS "MACHINE SPEECH". THE DEVELOPMENT OF THESE INNOVATIVE TECHNOLOGIES WILL BE BASED ON EMERSON & STERN'S EXISTING VOICE I/O CAPABILITIES FOR REAL-TIME SPEECH PROCESSING; CAREFUL ATTENTION WILL BE PAID TO THE HUMAN FACTORS RAISED BY USING VOICE. WORK TO BE DONE IN PHASE I INCLUDES COLLECTING AND ANALYZING EXISTING VOICE INTERACTIONS IN THE COCKPIT; DEFINING A "COMMAND" LANGUAGE FOR THE SAMPLE APPLICATION OF DATABASE INQUIRY AND RESPONSE; DETERMINING HUMAN PERFORMANCEON THIS COMMAND LANGUAGE AT VARIOUS NOISE LEVELS; ANALYZING SUPRASEGMENTAL DATA; DEVELOPING APPROPRIATE "MACHINE SPEECH"WITH PROPRIETARY TOOLS; ADAPTING EMERSON & STERN'S SPEECH ALGORITHMS TO COCKPIT NOISE CONDITIONS; AND FINALLY, COMPARING RECOGNIZER PERFORMANCE TO HUMAN PERFORMANCE AND THE INTELLIGIBILITY OF HUMAN SPEECH TO "MACHINE SPEECH".