Voice conversion model for estimation of transmission characteristics in auditory feedback
* Presenting author
We propose a voice conversion model for estimation of transmission characteristics in auditory feedback. Speakers feel sense of discomfort when they listen recorded speech sound of themselves. The phenomenon is caused by different transmission pathways. The recorded speech sound contains just air conducted sound from the speakers. On the other hand, the speech sound perceived by the speaker contains both air- and bone- conducted sounds in auditory feedback. A method for measuring transfer function and transmission characteristic in auditory feedback has never proposed, since direct observation of the sound in the inner ear is impossible. In addition, there is a problem that the sound source and listening positions cannot be underspecified. In this paper, we propose a voice conversion model from recorded speech sound for estimation of transmission characteristics in auditory feedback. The transmission characteristic in auditory feedback will be estimated by subtracting air conduction sound from converted voice. The converted voice was obtained by synthesizing air- conducted sound and bone- conducted sound with time delay, sound pressure level, and reverberation as parameters. Bone- conducted speech was obtained by filtering from air conducted sound.