Research on the theory and practice of multimedia music image photography in music research

Abstract

Note recognition is the core and key of music score recognition. In this paper, according to the diversity and polymorphism of notes, a structure-based recognition scheme is determined, and its recognition process is divided into two stages: primitive extraction and structure analysis. In the aspect of note primitive extraction, the rough extraction method based on vertical run-length coding and the fine detection method based on horizontal run-length coding are proposed, which overcome the shortcomings of the existing methods such as poor adaptability of complex notes and incomplete extraction results. A solid head extraction method is designed, which first divides and then detects the features. This method uses the prior knowledge of notes and the existing spectral lines and trunk recognition. In the aspect of note structure analysis, a new method of note structure analysis based on action field is proposed. In this method, the concept of action field in physics is introduced into the relationship expression of note primitives, and the unity of knowledge, generality and accuracy is realized. On this basis, six note substructures are defined, and a note structure analysis model of key structure priority location is established, which realizes the reconstruction of the note base metadata to the note object. This model reflects the thinking habit of highlighting key features and from the whole to the details in manual spectrum recognition. It not only reduces the complexity of analysis, but also has a strong ability of primitive redundancy error removal. The test results show that the overall recognition performance has reached the level of the current excellent commercial music score recognition system, and it has obvious advantages in note recognition, adaptability in different data environments and execution speed.