Analysis of artistic emotion expression and harp performance characteristics based on deep learning

Abstract

Art is the spice of life and another interpretation of life. Rich emotions make art more vital and appealing. In the art of performance, emotional expression is very important, which can not only enhance the expressive power of performance, but also sublimate the emotion contained in the work. In the eighteenth century, the harp was first played in opera orchestras. From the mid-nineteenth century, the harp was widely used in symphony orchestras. The harp became an important large-scale plucked instrument in modern orchestras. Composers are even more fond of her. No matter in solo or in band, the artistic charm and performance value of harp can not be replaced. Because of its unique charm of timbre, it is very infectious and unforgettable to the audience. This paper tries to analyze the artistic characteristics of this work more objectively, and discusses the creative ideas of combining the work with performance, so as to provide good experience for the performance practice of this work to a certain extent. At present, the recognition and classification system in the field of music information retrieval mainly extracts music features manually, then trains the model with the classifier, and finally uses the built model to recognize and classify the test music samples. At present, however, there is a bottleneck in extracting music features manually. As a new feature extraction technology, deep learning has achieved great results in image processing, natural language understanding and other fields. Therefore, this paper aims to use the powerful feature extraction ability of deep learning to find more suitable music features for music genres. And the problem of identification and classification of traditional Chinese musical instruments has been studied. In this paper, a harp recognition and classification algorithm based on deep learning is proposed. The music samples of each instrument are preprocessed and Mel frequency cepstrum coefficients are extracted, which are then input into the depth confidence network, and then adjusted for training. Finally, the trained model is used to predict the types of test instruments.