摘要 |
PURPOSE: An automatic labeling method of a cut-reading in case of a voice mixing is provided to automatically label a cut-reading of a voice DB(Data Base) without using a man's hand. CONSTITUTION: If a silence of a speech DB is not the next of a closed sound, a border of intonation phrase is displayed. If a silence of a speech DB is the next of a closed sound, a length of a silence of a voice DB is compared with a predetermined threshold value. If the length of the silence of the voice DB is longer than the predetermined threshold value, a border of the intonation phrase is displayed. If the length of the silence of the voice DB is shorter than the predetermined threshold value, a non-border of the intonation phrase is determined. Until the silence of the voice DB reaches to the last period, the step of displaying the border of the intonation phrase and the step of determining the non-border of the intonation phrase are repeated. |