SPEECH RECOGNITION APPARATUS

摘要

Speech processing apparatus has a store (30) containing a reference vocabulary (31) of words and carries out active word selection in accordance with mode data supplied to the apparatus. After this, the apparatus performs dynamic template adaptation in dependence on the output of a sensor (2) that monitors environmental influences on the speaker of the kind that modify speech sounds made by the speaker. The sensor (1) may be responsive to vibration or acceleration forces on the speaker, the apparatus using information about how vowel and consonant groupings are influenced by forces on the speaker to perform the template adaptaton. Pattern matching is used to compare the speech sounds of the speaker with the modified signals after template adaptation to identify the most likely word spoken.