ALTHOUGH MEMBERS OF THE LINGUISTIC AND SPEECH RECOGNITION COMMUNITIES RECOGNIZE THE IMPORTANCE OF SUPRASEMENTALS (DURATION, INTENSITY, AND FUNDAMENTAL FREQUENCY OR PITCH) INUNDERSTANDING THE SPEECH CODE, THERE IS RELATIVELY LITTLE QUANTIFIABLE DATA

ALTHOUGH MEMBERS OF THE LINGUISTIC AND SPEECH RECOGNITION COMMUNITIES RECOGNIZE THE IMPORTANCE OF SUPRASEMENTALS (DURATION, INTENSITY, AND FUNDAMENTAL FREQUENCY OR PITCH) INUNDERSTANDING THE SPEECH CODE, THERE IS RELATIVELY LITTLE QUANTIFIABLE DATA AVAILABLE, PARTICULARLY ON PITCH. EMERSON & STERN ASSOCIATES HAS ALREADY DEVELOPED A BASIC APPROACH FOR RELIABLY EXTRACTING PITCH INFORMATION FROM VARIOUS VOICES. THE OBJECTIVES OF PHASE I RESEARCH ARE TO COMPLETE AND VERIFY THE RELIABILITY OF THIS SOFTWARE TOOL; TO DETERMINE THE FEASIBLITY OF USING IT TO MANIPULATE INTONATINOAL CONTOURS; AND TO DETERMINE WHAT INFORMATION HUMAN LISTENERS CAN INTERPOLATE FROM "SUPRASEGMENTAL-ONLY" SYNTHESIZED SPEECH. THESE OBJECTIVES WILL BE ACHIEVED THROUGH THE EXECUTION OF VARIOUS EXPERIMENTS. PHASE II RESEARCH WOULD CONTINUE TO REFINE THE TOOL, EXTENDING ITS CAPABILITIES TO INCLUDE THE MANIPULATION OF DURATION, INTENSITY, AND PHONETIC COMPONENTS. SOFTWARE WILL BE DEVELOPED ON THE 512K MACINTOSH WITH THE INTENTION OF PROVIDING A RELATIVELY INEXPENSIVE TOOL FOR USEBY LINGUISTS AND OTHER SPEECH RESEARCHCERS. DATA GATHERED WITH SUCH A TOOL HAS IMPLICATIONS FOR LINGUISTIC RESEARCH AND TEH DESIGN OF AUTOMATIC SPEECH RECOGNIZERS, WITH MYRIAD POTENTIAL COMMERCIAL APPLICATIONS IN INDUSTRY, GOVERNMENT, LANGUAGE LEARNING, AND HANDICAPPED-USER SETTINGS.