| 摘要 |
PURPOSE: A device using PLU(Phoneme Like Unit) for registering a speaker independence/vocabulary independence voice model and a method thereof are provided to recognize vocabularies of infinite speakers by generating a general word model independent of a language characteristic of a specific speaker. CONSTITUTION: A device using PLU(Phoneme Like Unit) for registering a speaker independence/vocabulary independence voice model includes a voice recognition device(100), and a PLU extraction equipment(200). The voice recognition device(100) generates and stores a speaker independence, a vocabulary independence, and a word model corresponding to a voice model for registration in case that an operator inputs a voice model for registration, and includes a key board(101), a ROM(102), a RAM(103), and a CPU(104). The key board(101) outputs text data about the voice model for registration to the CPU(104) in case that the operator inputs the voice model for registration. The RAM(103) stores numerous word models corresponding to the numerous voice models for registration. The CPU(104) divides the voice model for registration about the text data into numerous PLU, and reads only PLU binary data from the ROM(102), and builds a speaker independence/vocabulary independence word model by using the PLU binary data. At this time, the PLU binary data are corresponding to the divided numerous PLU among numerous PLU divided according to a kind of speaker. Meanwhile, the PLU extraction equipment(200) converts the voice data into the divided PLU binary data by receiving the voice data inputted by numerous speakers, and stores the PLU binary data at the ROM(102). A PLU extraction part(203) of the PLU extraction equipment(200) receives digital voice data corresponding to each speaker from a DAT(Digital Audio Tape)(202), and extracts the divided PLU binary data from the digital voice data, and stores the divided PLU binary data at the ROM(102). |
