Application of Lightweight Deep Learning Model in Vocal Music Education in Higher Institutions

Abstract

The aim is to improve the teaching quality of music majors and cultivate their innovative ability. This article takes Vocal Music Education (VME) method as the research object to explore the teaching reform of Music Major courses. Firstly, this article makes an in-depth study on Big Data Analytics (BDA) and Digital Twins (DTs) technology and constructs a DTs platform connecting real teaching space and virtual teaching space. Secondly, the DTs platform is divided into online learning feature analysis and virtual-real teaching space integration functional modules. This article explores the online immersive education process design and technology application of the DTs platform from the two aspects of teaching and technology. Afterward, it designs a student action and expression recognition network based on the Visual Geometry Group (VGG) Net model and Google Net model in teaching data collection and management. Finally, the proposed system is tested. The test results show that the active and passive interaction curves of the traditional VME system have no obvious fluctuation, indicating that the interaction of the traditional VME system is not strong, and the ability of active feedback information is poor. By contrast, the active and passive interaction curves in the proposed VME have large fluctuations, showing that the proposed VME has more frequent interaction, and the teaching information can get real-time and active feedback. Therefore, the proposed VME system can better stimulate students' learning desire. Meanwhile, the constructed Neural Network (NN) has the highest recognition accuracy of 99.07% on the student action and expression dataset. When tested with the image data taken by the research experiment, the highest accuracy is 89%, with an average of more than 85%. The proposed VME system provides ideas for applying DTs technology in the college of music education.