PROBLEM TO BE SOLVED: To provide an apparatus, method and program for word spotting speech recognition which are not affected even when an unnecessary word model has no high precision and is not matched with a speech in an unnecessary word section.;SOLUTION: A feature parameter generation part which cuts a speech section out of an inputted speech and divides it into frames, an acoustic model storage part which stores feature parameters of the speech in unit of sub-word, a keyword model generation part which generates a keyword model and an unnecessary word model from reading data of a keyword and reading data of an unnecessary word, an unnecessary model generation part, a keyword likelihood calculation part, an unnecessary word likelihood calculation part, a limit value setting part which regards a set value as unnecessary likelihood when unnecessary word likelihood is smaller than the preset value, and a viterbi processing part which calculates cumulative likelihood, are included.;COPYRIGHT: (C)2004,JPO
展开▼