摘要
盲人音乐家在交流创作的音乐作品时面临着人工转换和效率较低的问题,信息科学与技术的迅速发展为解决此类问题提供了许多解决方案。虽然目前有许多盲文音乐作品的识别方案,但其存在识别效率低和兼容能力不足等缺点。为了避免传统方案在盲文音乐图片特征提取时过多依赖人工经验,通过研究提出并设计了基于卷积神经网络的识别模型。在对盲文音乐图片的样例数据进行预处理之后,通过多次反复迭代训练,模型就可学习到盲文音乐图片中音乐符号的特征。实验结果表明,该模型的识别有效性和较强的泛化能力为盲文音乐作品的识别提供了一种新的解决方案。
Blind musicians are confronted with the problems of manual conversion and low efficiency in the communication of musical works.The rapid development of information science and technology has provided many solutions to these problems.However,most of the recognition schemes for braille music works lack recognition efficiency and compatibility.In consideration of this deficiency,whereby traditional schemes rely heavily on artificial experience in braille music picture extraction,a convolution neural network-based recognition model has been developed.After preprocessing the sample data of braille music pictures through repeated iterative training,the recognition model was able learn the characteristics of music notation in braille music pictures.The experimental results showed the recognition effectiveness and strong generalization ability of the model,which provides a new idea for the recognition of braille music works.
作者
刘彪
黄蓉蓉
林和
苏伟
LIU Biao;HUANG Rongrong;LIN He;SU Wei(School of Information Science and Engineering,Lanzhou University,Lanzhou 730000,China;No.69230 Troops of PLA,Wusu 833000,China)
出处
《智能系统学报》
CSCD
北大核心
2019年第1期186-193,共8页
CAAI Transactions on Intelligent Systems
基金
广西科技计划项目(桂科AA17204096
桂科AD16380076)
兰州市人才创新创业科技项目(2014-RC-3)
关键词
机器学习
盲文音乐识别
卷积神经网络
深度学习
计算机视觉
图像识别
人工智能
图像处理
machine learning
braille music recognition
convolution neural network
deep learning
computer vision
image recognition
artificial intelligence
image processing
