| 摘要 |
Known speech recognition apparatus, for recognising several different spoken statements comprising several words, includes an acoustic recogniser that compares speech sounds from a microphone with signals stored in a reference vocabulary. When words making up a statement are recognised, a signal is supplied to control equipment associated with that statement. In the present invention the apparatus has a processing unit with groups of several storage locations. Each group is associated with a different statement, each location in that group being associated with an individual word. Signals representing the spoken words are used to mark those locations in every group that has a location associated with the word. An output unit identifies which group is most nearly filled with marked locations and hence produces an output indicative of the spoken statement which is supplied to the equipment associated with that statement. Where a group includes two or more locations associated with the same word, they are arranged serially so that only one location at a time is marked with the spoken word each time it is spoken. |
