| 摘要 |
PURPOSE:To improve the conversion efficiency to a proper character string by adding a language adaptation to voice adaptation and determining the display priority in case of displaying a display character string already matching with a dictionary. CONSTITUTION:When sentence data are inputted from a microphone 1 as voice, a voice feature parameter time series related to spectrum information or the like is extracted by a voice parameter extracting part 2. A voice recognition part 4 matches the extracted time series with a reference pattern stored in a voice dictionary 3 and stores a CV syllable proposed sequence which is a recognized result in a syllable lattice storage part 5 in accordance with proposed order. A conversional object selecting part 6 combines syllables to be proposed to form a rough character string finds out similarity based upon a distance from a reference pattern and transfers the similarity to a dictionary matching part 9. The matching part 9 retrieves a word dictionary 8 and stores proper characters in a storage part. A display selecting part 11 displays a character having the largest load sum between the voice adaptation and language adaptation in priority. The language adaptation is obtained from a format prime, word using frequency, etc. |
