PHONEME RECOGNITION SYSTEM

A PROGRAM OF RESEARCH IS PROPOSED TO ESTABLISH THE DESIGN PARAMETERS FOR A NOVEL "SEQUENTIAL PHONETIC SORTING MACHINE"TO BE USED IN AUTOMATIC SPEECH RECOGNITION. THIS NEW SIGNAL PROCESSING TECHNIQUE HAS THE SIMPLICITY, ROBUSTNESS, AND LOW COST THAT IS LACKING IN THE SYSTEMS NOW USED BY THE HANDICAPPED TO CONTROL THEIR ENVIRONMENT BY VOICE COMMANDS. VOIS, INC., HAS ANALYZED THE REQUIREMENTS FOR SIGNAL PROCESSING OF THE TOTAL ACOUSTIC ENVIRONMENT AS WELL AS THE CHARACTERISTICS OF HUMAN SPEECH AND HAS DETERMINED THAT THERE ARE INTRINSIC LIMITATIONS IN PRESENT SPEECH PROCESSORS ARISING OUT OF THEIR FUNDAMENTAL THEORETICAL BASIS AND METHODOLOGIES. THIS LIES IN THE USE OF THE FOURIER TRANSFORM, OR ANY LEAST SQUARE POLYNOMIALS, AS THE BASIS FOR SPEECH SIGNAL PROCESSING. VOIS HAS DEVELOPED AN ALTERNATIVE METHOD THAT USES THE PRINCIPLES OF "SORTING AND CLASSIFYING" THE WAVEFORM STRUCTURE. THIS SYSTEM HAS ALREADY DEMONSTRATED THE ABILITY TO RECOGNIZE VOWELS AND CONTINUOUS CONSONANTS FROM THE WAVEFORM AND USE THE INFORMATION TO SEGMENT SPEECH INTO PHONEMES. THE PROPOSED PROGRAM WILL CARRY OUT RESEARCH NEEDED TO RECOGNIZE ALL ENGLISH PHONEMES AND TO VERIFY THE FEASIBILITYOF USING THE CONCEPT IN AN ASSOCIATIVE PROCESSOR. VOIS' PROCESSOR WILL PROVIDE A LOW COST MEANS OF INPUT