DOI QR코드

DOI QR Code

Korean Traditional Music Melody Generator using Artificial Intelligence

인공지능을 이용한 국악 멜로디 생성기에 관한 연구

  • Bae, Jun (Department of Music Technology, The University of Suwon)
  • Received : 2021.03.07
  • Accepted : 2021.05.30
  • Published : 2021.07.31

Abstract

In the field of music, various AI composition methods using machine learning have recently been attempted. However, most of this research has been centered on Western music, and little research has been done on Korean traditional music. Therefore, in this paper, we will create a data set of Korean traditional music, create a melody using three algorithms based on the data set, and compare the results. Three models were selected based on the similarity between language and music, LSTM, Music Transformer and Self Attention. Using each of the three models, a melody generator was modeled and trained to generate melodies. As a result of user evaluation, the Self Attention method showed higher preference than the other methods. Data set is very important in AI composition. For this, a Korean traditional music data set was created, and AI composition was attempted with various algorithms, and this is expected to be helpful in future research on AI composition for Korean traditional music.

음악 분야에서는 최근 머신러닝을 이용한 다양한 인공지능 작곡 방법이 시도되고 있다. 하지만 이 연구는 대부분 서양음악을 중심으로 이루어져왔고 국악에 관한 연구는 거의 이루어지지 않았다. 특히 연구를 위한 데이터 세트조차 만들어지지 않은 상태여서 연구에 어려움이 많았다. 이에 해당 논문에서는 국악의 데이터 세트를 만들고 그 데이터 세트를 기반으로 하여 세 가지 알고리즘을 이용하여 국악 멜로디를 생성하고 그 결과물을 비교하여 보기로 한다. 언어와 음악의 유사성에 기반한 LSTM, Music Transformer 그리고 Self Attention 3가지 모델들이 선택되었다. 각 3가지 모델을 이용하여 국악 멜로디 생성기를 모델링하고 학습시켜 국악 멜로디를 생성해 내었다. 사용자 평가 결과 Self Attention 방식이 LSTM 방식과 Music transformer 방식에 비해 높은 선호도를 보였다. 데이터 표현 및 훈련데이터는 인공지능 작곡에 있어 매우 중요하다. 이를 위한 기초적인 국악 데이터 세트를 만들고 다양한 알고리즘으로 인공지능 작곡을 시도하였고 이것이 향후 국악 인공지능 작곡의 연구에 도움이 될 수 있을 것으로 기대한다.

Keywords

References

  1. L. Wyse. Mechanisms of artistic creativity in deep learning neural networks.arXiv preprint, arXiv:1907.00321, 2019.
  2. Y. Nam and Y. Kim, "Melody composition using geometric crossover for variable-length encoding," Computational Intelligence in Music, Sound, Art and Design Lecture Notes in Computer Science, pp. 37-38, 2019.
  3. G. Hadjeres, F. Pachet, and F. Nielsen, "Deepbach: a steerable model for bach chorales generation," Proceedings of the 34th International Conference on Machine Learning, PMLR 70, pp. 1362-1371, 2017.
  4. H. Hild, J. Feulner, and W. Menzel, "HARMONET: A neural net for harmonizing chorales in the style of JS Bach," Applications of Evolutionary Computation, pp. 267-274, 1992.
  5. B. Freisleben, "The neural composer: A network for musical applications," International Computer Music Association, vol. 1993, pp. 1663-1666, 1992.
  6. N. Boulanger-Lewandowski, Y. Bengio, and P. Vincent, "Modeling temporal dependencies in high-dimensional sequences: Application to polyphonic music generation and transcription," arXiv preprint, arXiv:1206.6392, 2012.
  7. R. Vohra, K. Goel, and J. K. Sahoo, "Modeling temporal dependencies in data using a DBN-LSTM," arXiv:1803.01271, pp. 1-4, 2015.
  8. S. Jee, "Comleteing Unfinished Classic Music Using AI," Seoul School of Intergrated Sciences and Technologies, vol. 1, no. 1, pp. 2-4, 2020.
  9. J. Bae, "Deep Learning Music Genre Classification System Model Improvement Using Generative Adversarial Networks (GAN)," International Journal of Information and Communication Engineering, vol. 24, no. 7, pp. 842-848, 2020.
  10. D. Bahdanau, K. Cho, and Y. Bengio, "Neural machine translation by jointly learning to align and translate," arXiv preprint, arXiv:1409.0473, 2014.
  11. J. Bae and C. Y. Kim, "Deep Learning Music genre automatic classification voting system using Softmax," International Journal of Information and Communication Engineering, vol. 23, no. 1, pp. 27-32, 2019.