基于深度学习的钢琴手型指法识别系统  

Piano Hand Fingering Recognition System Based on Deep Learning

作　　者：唐浩鑫 胡建华 魏嘉俊 黄沃清 张晓南 任楚 TANG Haoxin;HU Jianhua;WEI Jiajun;HUANG Woqing;ZHANG Xiaonan;REN Chu(Computer Engineering Technical College,Guangdong Polytechnic of Science and Technology,Zhuhai 519090,China)

机构地区：[1]广东科学技术职业学院计算机工程技术学院,广东珠海519090

出　　处：《现代信息科技》2023年第19期17-20,共4页Modern Information Technology

基　　金：广东省科技创新战略专项资金(“攀登计划”专项资金)(pdjh2023a0848);广东省教育厅青年创新人才类项目(2019GKQNCX043)。

摘　　要：随着音乐素质教育的普及,学习钢琴的儿童与日俱增。但由于家长工作忙、不懂专业知识、孩子不愿主动练琴等原因,需要专人进行“钢琴陪练”。目前行业内人工陪练存在收费价格昂贵、陪练人员不专业、无法进行可视化分析等缺点。因此,文章开发了基于深度学习的钢琴手型指法识别钢琴教学陪练系统,系统通过获取摄像头实时拍摄数据进行实时处理,通过目标检测、姿态识别等方式识别出手的关节点,与名师数据集进行比对,使用图像分类算法识别错误手型,将实时处理后的视频结果通过推流的方式,将识别结果在网页端展示,具有较好的实际应用价值。With the popularization of music quality education,the number of children learning piano is increasing day by day.However,due to busy parents'work,lack of professional knowledge,and children's unwillingness to take the initiative to practice piano,a dedicated person is needed to accompany them in piano practice.At present,there are shortcomings in the industry,such as expensive fees,unprofessional accompaniment personnel,and inability to conduct visual analysis for manual accompaniment.Therefore,this paper develops a piano teaching accompaniment system based on Deep Learning for piano hand fingering recognition.The system obtains real-time camera shooting data for real-time processing,identifies joint points of the hand through target detection,posture recognition and other methods,and compares them with the dataset of famous teachers.It uses image classification algorithm to recognize incorrect hand types,and pushes the real-time processed video results to display on the webpage.It has good practical application value.

关 键 词：深度学习 目标检测算法 姿态识别算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]