| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a speech processing device which generates an appropriate masker sound while preventing howling. <P>SOLUTION: A voice analysis part 13 extracts an amount of voice features such as formants, pitch, and the like from an input voice signal. A masker sound generation part 14 processes a sound source (a general purpose masker sound) stored in a database 15 based on the extracted amount of voice features to generate a masker sound for output. The general purpose masker sound is for example a recording of voices of multiple people including men and women and includes a disturbance sound having no lexical meaning (conversation contents being unintelligible). The masker sound generation part 14 matches the formant of the disturbance sound with the formant of the input voice signal. The generated masker sound has no lexical meaning and includes a disturbance sound approximated to a speaker in the voice quality and the voice tone, thereby making it difficult for a listener to understand the speech contents of the actual speaker. <P>COPYRIGHT: (C)2012,JPO&INPIT |
