Speech recognition method, speech recognition device, and recording medium on which is recorded a speech recognition processing program

Abstract

This invention concerns obtaining high recognition capability while there is a large limitation on memory capacity and processing ability of a CPU. When several words are selected as registration words among a plurality of recognizable words, a recognition target speaker speaks the respective registration words, registration word data for the respective registration words from the sound data is created and saved in a RAM. When the recognition target speaker speaks a registration word, sound is recognized using the registration word data, and when recognizable words other than the registration words are recognized, sound is recognized using specific speaker group sound model data. Furthermore, speaker learning processing is performed using the registration word data and the specific speaker group sound model data, and when recognizable words other than the registration words are recognized, sound is recognized using post-speaker learning data for speaker adaptation.

Dependencies are inferred from the claim text.