随着电子音乐制作的普及,乐器数字接口(Musical Instrument Digital Interface,MIDI)信号的有效转换对音乐创作的意义日益凸显.为了提高电子音乐旋律自动转换的准确率,研究采用了一系列深度神经网络模型,对大型乐器数字接口数据集进行学习,并利用 Adam 算法进行优化.结果显示,在训练阶段和测试阶段,模型的识别率分别达到了 95.03%和96.12%.这一结果不仅在训练集上超出了基线模型 16.88%的表现,在测试集上的提升更是达到了 18.36%,展现了模型优化后的强大泛化能力.在关键的性能指标上,模型的精确率为 94.03%,召回率为 95.06%,以及F1 分数高达 97.89%,这些指标均显著高于基线模型的原始得分.综合可以看出,基于深度学习的MIDI信号对电子音乐旋律自动转换研究,有效地提高了音乐转换的精度.
The Automatic Conversion of Electronic Music Melody by MIDI Signals Based on Deep Learning
With the popularization of electronic music production,the effective conversion of Musical Instrument Digital Interface(MIDI)signals has become increasingly important for music creation.In order to improve the accuracy of automatic melody conversion in electronic music,a series of deep neural network models were used to learn large instrument digital interface datasets,and Adam algorithm was used for optimization.The results showed that in the training and testing stages,the recognition rates of the model reached 95.03%and 96.12%,respectively.This result not only exceeded the performance of the baseline model by 16.88%on the training set,but also achieved an improvement of 18.36%on the test set,demonstrating the powerful generalization ability of the opti-mized model.In terms of key performance indicators,the accuracy of the model is 94.03%,the recall rate is 95.06%,and the F1 score is as high as 97.89%,all of which are significantly higher than the original score of the baseline model.Overall,it can be seen that the research on electronic music melody automatic conversion based on deep learning MIDI signals has effectively improved the accuracy of music conversion.
deep learningMIDI signalmusic melodyautomatic conversionadam algorithm
万方数据
维普
