基于改进特征提取方法的五线谱识别
|
摘要
针对传统五线谱识别方法存在谱线过删和漏删的缺点,以及现有音符特征提取方法与谱线删除相互制约的问题,提出一种改进的、无需删除谱线的特征提取方法。在图像预处理阶段保留谱线,将音符与谱线同时投影,结合音符符杆垂直像素数据与音符其它部位像素数据携带的映射特征,对横纵向投影数据进行数理统计分析,得到供音符类型识别的归一化特征值,再利用基准谱线与音符符头的相对位置获取音调信息。实验结果表明,该方法在保证较高识别精度的基础上,进一步提高了识别速率,可以有效识别音符组合形式较复杂的乐谱,对于五线谱识别应用具有重要意义。
Concerning the excessive-deleting problem and leaky-deleting problem that existed in spectral delete of traditional recognition of music score,and the mutual restraint between the feature extraction algorithm of notes and the lines delete,a novel feature extraction algorithm is proposed,which preserves the spectral lines. We preserved the spectral lines in image preprocessing and projected the notes and spectral lines at the same time. Combined with the mapping characteristics carried by the vertical pixel data of notes rod and pixel data of other parts of notes,performing basic mathematical statistical analysis on the horizontal and vertical projection data,the normalized feature values for musical note type recognition are obtained. Then we used the relative location between the standard lines and notes head to obtain the tone information. The results of the simulation illustrate that the proposed algorithm is a feasible way to recognize more complex combinations of notes and improve the recognition rate under the premise of ensuring high recognition accuracy.
Concerning the excessive-deleting problem and leaky-deleting problem that existed in spectral delete of traditional recognition of music score,and the mutual restraint between the feature extraction algorithm of notes and the lines delete,a novel feature extraction algorithm is proposed,which preserves the spectral lines. We preserved the spectral lines in image preprocessing and projected the notes and spectral lines at the same time. Combined with the mapping characteristics carried by the vertical pixel data of notes rod and pixel data of other parts of notes,performing basic mathematical statistical analysis on the horizontal and vertical projection data,the normalized feature values for musical note type recognition are obtained. Then we used the relative location between the standard lines and notes head to obtain the tone information. The results of the simulation illustrate that the proposed algorithm is a feasible way to recognize more complex combinations of notes and improve the recognition rate under the premise of ensuring high recognition accuracy.
引文
[1]BLOSTEIN D.Using diagram generation software to improve diagram recognition:a case study of music notation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1999,21(11):1121-1136.
[2]ROSSANT F,BLOCHB I.A fuzzy model for optical recognition of musical scores[J].Fuzzy Sets and Systems,2003,141(2):165-201.
[3]ROSSANT F.A global method for music symbol recognition in typeset music sheets[J].Pattern Recognition Letters,2002,23(10):1129-1141.
[4]MIYAO H,NAKANO Y.Note symbol extraction for printed piano scores using neural networks[J].IEICE Transactions on Information and Systems,1996,E79-D(5):548-554.
[5]FAHMY H.A graph-rewriting approach to discrete relaxation application to music recognition[C].Proceedings of SPIE,1994:291-302.
[6]张登盛,马颂德.多声部乐谱图像识别系统[J].计算机工程,1998,24(3):41-43.
[7]王林泉,章文怡.区域特征的乐谱识别系统[J].软件学报,1994,5(11):44-49.
[8]RANDRIAMAHEFA R,COCQUEREZ J P,FLUHR C,et al.Printed music recognition[C].Proceedings of the 2nd IEEE International Conference on Document Analysis and Recognition,1993:898-901.
[9]BAINBRIDGE D,BELL T C.Dealing with superimposed objects in optical music recognition[C].Proceedings of the 6th IET International Conference on Image Processing&Its Applications,1997:756-760.
[10]GEORGE S E.Visual perception of music notation:on-Line and off line recognition[M].New York:IGI Publishing,2004.
[11]BAINBRIDGE D.Extensible optical music recognition[D].Christchurch:University of Canterbury,1997.
[12]FUJINAGA I,PENNYCOOK B,et al.The optical music recognition project[J].Computers in Music Research,1991(3):139-142.
[13]胡晓宏.基于链码特征的几何图形快速识别算法[J].吉林大学学报:理学版,2015,53(3):490-491.
[14]IICARINEN J.A shape recognition of irregular objects[C].Proceedings of SPIE the International Society for Optical Engineering,1996:1-8.
[15]刘清,林土胜.基于数学形态学的图像边缘检测算法[J].华南理工大学学报:自然科学版,2008,36(9):113-116.
[16]GERAUD T.A morphological method for music score staff removal[C].IEEE International Conference on Image Processing,2015:2599-2603.
[17]MONTAGNER I S,HIRATA R,HIRATA N S T.Learning to remove staff lines from music score images[C].IEEE International Conference on Image Processing,2014:2614-2618.
[18]陈皓,马彩文等.基于灰度统计的快速模板匹配算法[J].光子学报,2009,38(6):1587-1588.
[19]KOBAYAKAWA T.Auto music score recognizing system[C].Proceedings of SPIE:Character Recognition technologies,1993:112-123.
[20]MARTIN P,BELLISSANT C.Low-level analysis of music drawing images[C].First International Conference on Document Analysis and Recognition,1991:417-425.
[21]CHANG S,SOAK S.Optical music recognition using skeleton structure and neural network[C].Proceeding of SPIE,2002,4734:132-140.
[22]刘晓翔.光学乐谱识别技术研究与实现[D].西安:西北工业大学,2006,76-87.
[2]ROSSANT F,BLOCHB I.A fuzzy model for optical recognition of musical scores[J].Fuzzy Sets and Systems,2003,141(2):165-201.
[3]ROSSANT F.A global method for music symbol recognition in typeset music sheets[J].Pattern Recognition Letters,2002,23(10):1129-1141.
[4]MIYAO H,NAKANO Y.Note symbol extraction for printed piano scores using neural networks[J].IEICE Transactions on Information and Systems,1996,E79-D(5):548-554.
[5]FAHMY H.A graph-rewriting approach to discrete relaxation application to music recognition[C].Proceedings of SPIE,1994:291-302.
[6]张登盛,马颂德.多声部乐谱图像识别系统[J].计算机工程,1998,24(3):41-43.
[7]王林泉,章文怡.区域特征的乐谱识别系统[J].软件学报,1994,5(11):44-49.
[8]RANDRIAMAHEFA R,COCQUEREZ J P,FLUHR C,et al.Printed music recognition[C].Proceedings of the 2nd IEEE International Conference on Document Analysis and Recognition,1993:898-901.
[9]BAINBRIDGE D,BELL T C.Dealing with superimposed objects in optical music recognition[C].Proceedings of the 6th IET International Conference on Image Processing&Its Applications,1997:756-760.
[10]GEORGE S E.Visual perception of music notation:on-Line and off line recognition[M].New York:IGI Publishing,2004.
[11]BAINBRIDGE D.Extensible optical music recognition[D].Christchurch:University of Canterbury,1997.
[12]FUJINAGA I,PENNYCOOK B,et al.The optical music recognition project[J].Computers in Music Research,1991(3):139-142.
[13]胡晓宏.基于链码特征的几何图形快速识别算法[J].吉林大学学报:理学版,2015,53(3):490-491.
[14]IICARINEN J.A shape recognition of irregular objects[C].Proceedings of SPIE the International Society for Optical Engineering,1996:1-8.
[15]刘清,林土胜.基于数学形态学的图像边缘检测算法[J].华南理工大学学报:自然科学版,2008,36(9):113-116.
[16]GERAUD T.A morphological method for music score staff removal[C].IEEE International Conference on Image Processing,2015:2599-2603.
[17]MONTAGNER I S,HIRATA R,HIRATA N S T.Learning to remove staff lines from music score images[C].IEEE International Conference on Image Processing,2014:2614-2618.
[18]陈皓,马彩文等.基于灰度统计的快速模板匹配算法[J].光子学报,2009,38(6):1587-1588.
[19]KOBAYAKAWA T.Auto music score recognizing system[C].Proceedings of SPIE:Character Recognition technologies,1993:112-123.
[20]MARTIN P,BELLISSANT C.Low-level analysis of music drawing images[C].First International Conference on Document Analysis and Recognition,1991:417-425.
[21]CHANG S,SOAK S.Optical music recognition using skeleton structure and neural network[C].Proceeding of SPIE,2002,4734:132-140.
[22]刘晓翔.光学乐谱识别技术研究与实现[D].西安:西北工业大学,2006,76-87.