An online handwritten music symbol recognition system

Abstract

The objective of this study is to produce a system that would allow music symbols to be written by hand using a pen-based computer that would simulate the feeling of writing on sheets of paper and that would also accurately recognize the music symbols. To accomplish these objectives, the following methods are proposed: (1) Two features, time-series data and an image of a handwritten stroke, are used to recognize strokes; and (2) The strokes are combined, as efficiently as possible, and outputted automatically as a music symbol. As a result, recognition rates of 97.60% and 98.80% were obtained in tests with strokes and music symbols, respectively.

Online recognition of handwritten music symbols

“…In some studies such as [8], the offline methods were applied to images of music symbols right after they have been drawn. Meanwhile, Miyao and Maruyama proposed a novel penbased music symbol recognition system in [9]. Motivated from the fact that music symbols can be divided into common basic strokes, the system consists of two parts, one of which is to classify each input handwritten stroke as one of the basic strokes and the other is to combine the classification results into music symbols.…”

“…1. It has a similar structure as [9], but users of the proposed method do not have to be aware of any writing rules. To achieve the goal, the proposed method utilizes three types of features: the size information (SI), the histogram of directional movement angles (HDA), and the histogram of undirected movement angles (HUA).…”

“…To achieve the goal, the proposed method utilizes three types of features: the size information (SI), the histogram of directional movement angles (HDA), and the histogram of undirected movement angles (HUA). The original size of a stroke, which was used in [9] only to recognize the stroke for a dot, is important to distinguish other strokes as well without any prior knowledge. It is also noticeable that the additional performance improvement can be obtained by assigning more weights to the SI features than to the HDA and HUA features.…”

“…It is also noticeable that the additional performance improvement can be obtained by assigning more weights to the SI features than to the HDA and HUA features. The HDA features contain a time-series information, which is useful in handwriting recognition and is encoded by the chain code as in [9]. Especially, the HDA features are inspired by the descriptor of the scale invariant feature transform [10] and the histogram of oriented gradients (HOG) features [11], both of which are popular in various computer vision applications.…”

Recognition of pen-based music notation with finite-state machines

“…Miyao & Maruyama (2004, 2007 proposed a system based on predefined stroke primitives (such as note-heads, lines, dots, etc.). The combination of the time-series data and image features were employed to recognize strokes using Support Vector Machines.…”