| 摘要 |
PURPOSE:To emphasize an important word (key word) such as the article name and quantity of an order being voiced in, for example, telephone shopping (order reception). CONSTITUTION:An input voice is digitized and a feature extraction part 13 extracts a rhythm feature quantity and a spectrum feature quantity; and a key word extraction part 14 compares a time series of extracted spectrum features with spectrum feature time series previously registered in a key word dictionary 15 to detect the section of a matching key word, and an emphasizing process part 16 inserts pauses in front of and behind the detected key word section in the input voice, makes the pitch in the key word section, and increases the power, and then supplies the input voice to a voice synthesis part 13, which synthesizes and outputs a voice as a voice output. |
