Speech recognition in computer environment

SPEECH RECOGNITION IN COMPUTER ENVIRONMENT(M.Sc. Thesis)Abdulkadir KARACIGAZ UNIVERSITYINSTITUTE OF SCIENCE AND TECHNOLOGYSeptember 2006ABSTRACTIt has been for many years since the studies on ?Speech Recognition? were firstconducted. When these studies are examined, it is seen that most of the studies arebased on English language structure. The number of studies on ?Speech recognition?which are based on Turkish language structure is very limited.In this study, a software which recognizes the commands given in Turkish by aspeaker has been developed on Windows XP operating system. ?Pattern Recognition?method has been used in the software developed. To determine the beginning and theend of a word uttered, the ?Speech Recognition software? benefits from the ?energy ofthe sound?. Filtering, windowing, FFT, and auto-correlation processes are successivelyapplied to the sound which is caught. In this way, the ?reference signal? is receivedand, then, recorded to the database. When the user utters the same word again, theword is compared with the reference signals, and the recognition process ends. Theexperiements done on the software during the trail period show that such factors as thedistance from the microphone, the noise in the environment, the difference orsimilarity in the way a word is uttered, and the specifications of the sound-card and ofthe microphone affects the ?Speech Recognition? process.Science Code : 702.1.014Key Words : Speech recognition, pattern recognition, hidden markov modelPage Number : 134Adviser : Assistant Prof. Dr. Halil brahim BÜLBÜL