MACHINE INTELLIGENCE: IMPROVING SPEECH RECOGNITION WITH DYNAMICAL-NEURAL NETWORKS

OUT OF THE MANY PARAMETERS OF SPEECH RECOGNITION BY MACHINE INTELLIGENCE THERE IS ESPECIALLY ONE KEY ELEMENT THAT MUST HAVE A SOLUTION. THIS INVOLVES THE VARIABILITY OF SPEECH. WE HAVE PROPOSED THAT THE RECENTLY DEVELOPED DYNAMICAL-NEURAL NETWORK IS A CLASS OF MACHINE INTELLIGENCE WHICH HAS PROPERTIES THAT WARRANTS FURTHER RESEARCH AND DEVELOPMENT FOR CONTRIBUTING SOLUTIONS TO THE PROBLEMS OF SPEECH RECOGNITION. IN PARTICULAR, DYNAMICAL-NEURAL NETWORKS SHOW OPTIMISTIC PROMISE BECAUSE THEY ARE INHERENTLY FAST, THE NATURE OF THEIR FUNCTIONS SUPPORT GENERALIZATION, AND THEY SHOW POTENTIAL FOR STORING LARGE AMOUNTS OF MEMORY ASSOCIATIVELY. THE ONE AREA OF DEVELOPMENT FOR DYNAMICAL-NEURAL NETWORKS, WHICH WOULD MAKE A LARGE IMPROVEMENT IN THE APPLICATION OF SPEECH RECOGNITION, IS THE LEARNING ALGORITHM. IN THIS DOCUMENT WE DESCRIBE A NEW LEARNING ALGORITHM, BASED ON THE NEURAL MECHANISMS OF LEARNING IN THE HUMAN BRAIN, WHICH WE BELIEVE HAS PROMISE FOR CONTROLLING THE TYPE OF ASSOCIATIVE MEMORY REQUIRED TO MEET THE MANY REQUIREMENTS OF A SPEECH RECOGNITION SYSTEM. IT IS THE PURPOSE OF THE PROPOSED WORK TO TEST THIS ALGORITHM, IN PHASE I, THEN MAKE ADAPTIONS AS THE TESTS INDICATE; AND, IN PHASE II, APPLY THE RESULTING IMPROVED NETWORK TO SPEECH RECOGNITION FOR DEFENSE AND COMMERCIAL APPLICATIONS. ???????????