[様式－学 5]
Abstract of Doctoral Thesis
Title: A Study of a Highly-Realistic Singing-Experience System Based on Real-Time Vocoder and Sound Field Reproduction
Doctoral Program In Information Science and Engineering Graduate School of Information Science and Engineering Ritsumeikan University
ふりがな なかの こうた 氏 名 NAKANO Kota
Karaoke is one of the most popular forms of entertainment. Karaoke is singing, and it provides everyone opportunities for enjoyment. Karaoke is often used as a tool for communication, however, singing is also a skill. Some people, such as tone-deaf people, tend to avoid singing Karaoke with their companions. This tendency prevents good communication for them. In this thesis, I propose a system to solve the problem. The proposed system provides virtual singing-experience to users. The system modifies the singing-style of the users' singing-voice to professional-like one in real-time. The system also reproduces sound fields such as performance halls. Accordingly, the system provides users with good singing-experiences. The system supports singing for tone-deaf people, and also provides more entertainment than regular Karaoke to other users.
Firstly, I propose a singing morphing system by using vocoder framework based on the source-filter model. The system transcripts the singing-style of professional singers to amateur users’ singing-voice in real-time. To achieve the system, I propose an approach for a high quality vocoder, STRAIGHT to rapidly process the singing-voice, which depends on inverse-filtering method with STRAIGHT spectrum. According to the evaluations, I confirm that the proposed system can transcript singing-style of professional singers to users' singing-voice in real time.
Secondly, I propose a sound-reproduction system by using a semi-transaural loudspeaker-system and improved sound-field simulator. The semi-transaural
loudspeaker-system achieves high robustness for crosstalk of loudspeaker-system. The sound-field simulator depends on the finite-difference time-domain method. I propose an improvement for the simulator by employing spectral method and coordinate conversion for accurate computation. According to the evaluations, I confirm that the proposed system can robustly present fine sound-images to users.
Finally, I propose a singing-experience system with high-realistic sensation. The system depends on the integrated systems which is proposed in the previous sections in this thesis.
According to the evaluation, the sound-field simulator complemented the quality of singing morphing system. It was indicated that the integrated-systems could effectively provide good singing-experiences to users.
