Research On Algorithm Of Speaker-independent Isolated-words Recognition

Speech recognition is a cross-subject,which relates to signal processing, computer technology and mathematical modal, and is becoming a crucial interface technology between human and machine. It's application has been a rising and competitive high technology industry.One branch of speech recognition is speaker-independent isolated-word recognition(SIIWR).WIIWR technology got a quick development during the past years being applied in interaction between human and machine,intelligent phone,remote control of home electric equipment.WIIWR contains many crucial technologies including model-design, parameter abstraction and optimization, which will be discussed in this paper.Tasks of this paper are arranged by the following:To Study and describe key technology of voice signal used for speech recognition systematically and to verify it by experiments.To introduce the principle of wavelet transform and its application in speech recognition, and to focus on wavelet theory which is used in this paper, and to verify it by experiments.To study the principle of Hidden Markov Model and its three basic problems and the solutions to them, and it's application in speech recognition.Finally, to design a small vocabulary speaker-independent isolated-word recognition system and realize it in Matlab, and do some experiments to check it.