THE APPLICATION OF TRANSFORM IMAGING TO SPEAKER INDEPENDENT CONTINUOUS SPEECH RECOGNITION

SPEAKER INDEPENDENT, CONTINUOUS SPEECH RECOGNITION BY COMPUTERS HAS BEEN PURSUED FOR OVER THIRTY YEARS. CURRENTLY, THE MOST SUCCESSFUL VOICE RECOGNITION SYSTEMS RELY ALMOST EXCLUSIVELY ON HARDWARE SOLUTIONS AND DESPITE EXTREMELY POWERFUL MACHINES AND ELABORATE AND THOROUGH PROGRAMMING, THEY REQUIRE CONSIDERABLE TRAINING AND CUSTOMIZATION ON THE PART OF USERS. THIS PROJECT PROVIDES DESIGN DOCUMENTS FOR AND A PROOF-OF-CONCEPT SOFTWARE SYSTEM THAT RECOGNIZES SPEAKER INDEPENDENT AND CONTINUOUS SPEECH. THE AUTHORS TAKE A DIFFERENT APPROACH TO THE PROBLEM BY USING TWO NEW TECHNIQUES. ONE IS A PROPRIETARY WAY TO CAPTURE AND ANALYZE ACOUSTIC AND SEMATIC DATA THAT TARGETS THE INFORMATION HUMAN BEINGS USE IN SPEECH RECOGNITION. THE OTHER IS A PROPRIETY PATTERN MATCHING TECHNOLOGY CALLED "TRANSFORM IMAGING" THAT RAPIDLY MATCHES PATTERN ON THE BASIS OF THEIR SIMILARITY TO CHARACTERIZING EXAMPLES.