| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide voice recognition of all languages, and a single letter input method utilizing voice recognition. <P>SOLUTION: The invention relates to voice recognition of all languages, and the single word input method utilizing voice recognition, wherein a single letter is a letter of English, Chinese or the other languages, and pronunciation of the single letter is called a single tone. At first, m-pieces of unknown (known is possible) different single tones of one group represent m-pieces of different categories, a common single letter is pronounced once, and pronunciation of each single letter is converted to one linear prediction cepstrum coefficient matrix. Then, the common letter is classified into one category in m-categories, by using Bayes or distance classification method. After pronunciation of the single tone, a user searches F-pieces of unknown single tones which are most similar in the m-pieces of unknown single tones. Then, all single letters are arranged according to a similarity degree with a necessary single tone pronunciation and a letter (or a stroke counter), in an F-pieces group which is represented by F-pieces of similar unknown single tones. After pronunciation of the user, a necessary letter is very quickly searched. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
