Rechnereinheit zur Spracherkennung und Verfahren zur rechnergestützten Abbildung eines digitalisierten Sprachsignals auf Phoneme

摘要

The computer unit has an arrangement (MUS) for speech recognition independent of the speaker. A phoneme record memory (PS) stores phonemes and/or phoneme characteristic vectors for phonemes which have any desired number of segments. The vectors each characterise a phoneme. A user memory (BS) stores phoneme trains of digitised words, or digitised words spoken into the computer by the user. The phoneme trains are derived from a device which compares user characterising vectors with the stored phoneme vectors. The comparator (MV) compares user vectors of part of a digitised speech signal, which characterise that part of the signal, with the stored phoneme vectors. To form digitised speech signals from phonemes, the signal is divided into a number of digital speech signal components. For each component, a user characterising vector is derived which characterises the signal component. It is compared to stored phoneme vectors to provided a similarity value. The value is stored and in dependence on the value a train of phonemes is derived for the digitised speech signal from which the signal is composed.