Music Score Recognition and Composition Application Based on Deep Learning

Abstract

Optical score recognition is a critical technology for retrieving music information, and note recognition is a critical component of score recognition. This article evaluates and discusses the current state of research on important technologies for score recognition. To address the issues of low note recognition accuracy and intricate steps in the present music score image, a deep learning-based music score recognition model is proposed. The model employs a deep network, accepts the entire score image as input, and outputs the note's time value and pitch directly. Experiments on music scores demonstrate that the method described in this study has a high note identification accuracy of 0.95 for time values and 0.97 for pitch, which is suitable for composition.

Music Deep Learning: Deep Learning Methods for Music Signal Processing—A Review of the State-of-the-Art

“…For music score recognition, in [126], the proposed architectures takes as input an image of a music score, and outputs the duration, pitch, and coordinate for each note. Data from Muse Score [127] were used for the experiments, and the [128] model outperforms other architectures, with respect to all accuracy measures.…”

Classification of guitar tab and numbered musical notation using ResNet50 network

“…In order to solve the problem of long recognition time of music score images, Wu Qiong et al [7 ] improved CNN to residual CNN based on the C-BiLSTM model of convolutional neural circulation network and formed the residual convolutional neural circulation network RC-BiLSTM and recognized the optical music score based on the note properties of the music score. Liang Mingheng [8] uses the depth network to identify musical notes, accepts the entire musical score image as input, and outputs the time value and pitch of the notes.…”