Transferring singing expressions from one voice to another

Abstract

This paper presents an audio signal processing system that automatically transfers singing expressions from one voice to another. Depending on singers' skills, a song is sung with great variations in terms of note onset time, pitch and energy. The system focused on extracting and transferring musical expressions, excluding the timbre of singers. This singing expression transfer system can provide more intuitive guidance to those who want to learn new vocabulary expressions and help the music activities of those who have difficulty in singing. The system transfers expressions in the order of tempo, pitch, and energy. In this study, we propose an algorithm to align the tempo of the note, a method to match pitch and energy information, and a method to optimize the performance of these processes. Based on these methods, we propose a new singing expression transfer system and propose a new approach to singing voice modification.