摘要 |
PURPOSE:To realize extremely accurate voice-''kanji'' code conversion by obtaining a nondistortion fidelity voice signal, by mixing and multiplexing signals from microphones, differing in frequency characteristics, via filters. CONSTITUTION:As a voice input person reads a statement aloud in front of microphones 1a, 1b and 1c differing in frequency characteristics, inputs from the microphones are passes through band-pass filters 5a, 5b and 5c and mixed and multiplexed by multiplexer 6 and the feature parameter is extracted by parameter extractor 10 and inputted to voice discriminator 30. Voice discriminator 30 compares stored voice parameters with the feature parameter and inputs a voice code, which corresponds to a voice parameter of the greatest number of coincidence, to ''kana''- ''kanji'' converter 40 and from Japanese dictionary memory device 50, described word data corresponding to an index word equivalent to a row of voice is accessed and displayed, so that a selected ''kanji'' code is stored. |