Speech recognition method for use with a multimedia terminal, such as an interactive TV, whereby if a command sounds as if it could be one of two commands, the one previously selected most often is chosen

摘要

Method has the following steps: acquisition of a spoken command (2.3), processing of which generates two possible command candidates; comparison of probability values for a command being the correct one and, where the difference in probability values is less than a given threshold; execution of a decision step that involves comparison of the number of executions of the two commands and then selection of the command that has been executed most often. The invention also relates to a corresponding multimedia terminal.