Sheet Music Vocal Generator (SMVG)

Abstract

"Singing voice synthesis systems stand as remarkable achievements in modern technology. However, there persists a drive for improvement. To address this, our research proposes a novel approach: integrating sheet music—a traditional means of musical communication—as the input medium. By leveraging Optical Music Recognition (OMR), we aim to automate the detection and merging of essential musical elements like pitch, rhythm, and lyrics. This promises to enhance the efficiency and accuracy of singing voice synthesis processes significantly. The conventional methods for input in singing voice synthesis often involve manual transcription or recording, presenting inefficiencies and potential errors. By utilizing sheet music, a standardized format for musical notation, our approach aims to streamline the input process while ensuring precision in synthesis output. Incorporating OMR technology enables automated extraction and interpretation of musical data from sheet music, thereby enhancing synthesis efficiency. With advanced machine learning techniques, the model will recognize and extract pitch, rhythm, and lyrics with high accuracy, facilitating automated input and improving overall synthesis quality. Sheet Music Vocal Generator (SVMG) has proven its successfulness with its high accuracy classifications using a CRNN model."