Gladunov S. HARDWARE-SOFTWARE BASED TOOLS OF SEPARATED PHONEME LOCALISATION IN HUMAN AND COMPUTER SPEECH INTERACTION SYSTEMS

Research object there are hardware and software facilities of speech signal recognition. A research aim is to rise a flexibility of tuning the system of vocal human-computer interaction on the arbitrary phonetic composition of dictionary. Methods of theory of vocal pattern, theory of artificial neuron networks, theory of planning digital computing facilities, recognition are used. An apparatus are personal computers and FPGA Virtex XCV1000-FG680-4. Theoretical results: structural decomposition of the phoneme recognition module in speech recognition system by mean of new method of separate phoneme localization development is executed. New hardware structures of vocal computer management for acceleration of phoneme localization are developed. A rational structural organization of the phoneme localization hardware module is offered with hardware implementation on FPGA. Practical results: a rise of efficiency of vocal human-computer interaction systems exploitation by mean of flexibility rise and decreasing tuning time for the phoneme recognition module for the concrete dictator. Results are applied in the educational process in DonNTU and DonSMU, and also in the Donetsk regional clinical union. An area of application is a computing engineering.