抄録
Personal information included in the conversation at counters of hospitals or banks can be heard by a third party, and it is called 'Speech Privacy' problem. A sound masking technique is used against this problem. Though band limited pink noise is used in conventional masking, it causes loud noise for perfect masking because of its excess ingredient. Therefore, in this paper, we reduce excess ingredient included in masker by using linear predictive coding (LPC) to carry out the voice masking with lower sound pressure level. We confirm that the LPC based masker is more effective than the conventional masker against Japanese vowels by psychological experiments.
