| 摘要 |
<p>The invention relates to a voice recognition method with automatic correction in voice recognition systems with limited syntax. The inventive method comprises a word treatment stage (13) delivering a signal in a compressed form; a pattern recognition stage (14) in order to search for a phrase of a syntax which is the closest to said signal in compressed form on the basis of a syntax (SYNT1) formed by a set of phrases representing all possible paths between a set of words which were pre-recorded during a prior phase; storing (16) said signal in a compressed form; generating (17) a new syntax (SYNT2) wherein the path corresponding to the phrase determined during the previous recognition stage is prohibited; reiteration of the pattern recognition phase in order to search for another phrase which is closer to the stored signal on the basis of said new syntax.</p> |
