日本語音声スペクトルの特徴分析および音声認識・話者認識への考察

概要

論文の詳細を見る
Statistical analysis were made of static features of voiced sound spectra and threedimensional representation of consonants on the basis of the dynamic features of spectra. The spectral change of consonants, caused by the factor of speaker, vowel and time interval were also investigated and the following significant facts were found from these analysis. (1)Voiced consonant recognition for a limited speaker is easier than vowel recognition for unlimited speakers. (2)The relationship among consonants on the spectral space corresponds to the relationship in articulation. (3)The spectra of /i/, /u/, consonants followed by /i/ or /u/, and nasal are subject to the influence of the speaker-factor more than others for a short time interval between test and reference samples. (4)There is no significant difference of speaker-factor among consonants for a long time interval. (5)Long-term-independent personality in spectrum is extracted more easily from vowel spectrum than others
社団法人日本音響学会の論文
1979-03-01