SPEECH RECOGNITION DEVICE

摘要

A speech recognition device which can preferably be used for reducing the memory capacity required for recognizing speech of an unspecified person. A matching unit (114) spreads a speech model belonging to a first speech model network (300) and a garbage model (350) on an RAM (112) and supplies a speech parameter extracted by a speech parameter extraction unit (106) to the speech model of the RAM (112). Accompanying this, if an occurrence probability output from the garbage model (350) exceeds a predetermined value, a speech model belonging to any of the speech model group (402) is spread on the RAM (112) according to the occurrence probability output from the speech model belonging to the first speech model network (300).