Audio music feature recognition has been an active research area due to its various applications like music instrument classification. However, there is little literature investigating the possibility of classifying acoustic guitar strumming. This study aimed to classify acoustic guitar strumming in downstroke and upstroke style category, utilizing two neural network based classifier models. Each audio data were manually segmented per strum, and subjected to noise reduction through audio signal emphasis, downsampling, and MFCC conversion, before being fed to the designed neural network models. Convolutional Neural Network (CNN) and Long-Short-Term-Memory (LSTM) Recurrent Neural Network were separately used to classify the guitar strums. Results showed that CNN acquired a higher performance than LSTM with 91% model accuracy and a 19% model loss, while the LSTM model resulted in an 88% model accuracy and a 26% model loss. This indicates that CNN would be the preferred neural network model to utilize for classifying guitar strums. Future works may explore classification of other guitar types and the extraction of guitar strums in an audio file that includes other musical instruments.
Keywords: audio music feature recognition, guitar strumming style, guitar strum classifier[1] Mazhar, F. (2011). Automatic Guitar Chord Detection. Master’s Thesis, Tampere University of Technology, Tampere, Finland. Retrieved from https://trepo.tuni.fi/bitstream/handle/123456789/20923/mazhar.pdf?sequence=7&isAllowed=y
[2] Mistler, E. (2017). Generating Guitar Tablatures with Neural Networks. Master’s Thesis, University of Edinburgh, Edinburgh, Scotland. Retrieved from https://pdfs.semanticscholar.org/7065/897be10f94edb533ad8b98364372a85106da.pdf?_ga=2.81674903.1404172709.1544554413-1380252528.1544554413
[3] Matsushita, S. & Iwase, D. (2013). Detecting Strumming Action while playing Guitar. In Proceedings of the 2013 International Symposium on Wearable Computers, 145-146. ACM. Retrieved from: https://dl.acm.org/citation.cfm?doid=2493988.2494345
[4] Kittleberger, B. (2018, March 02). Acoustic Guitar Strumming Patterns. Retrieved from https://www.guitarworld.com/acoustic-nation/acoustic-guitar-strumming-patterns
[5] Uhlich, S., Giron, F., & Mitsufuji, Y. (2015, April). Deep Neural Network Based Instrument Extraction from Music. In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2135-2139. IEEE. Retrieved from https://ieeexplore.ieee.org/abstract/document/7178348
[6] McKay, C., & Fujinaga, I. (2005, March). Automatic Music Classification and the Importance of Instrument Identification. In Proceedings of the Conference on Interdisciplinary Musicology, 1-10. Retrieved from https://www.researchgate.net/profile/Ichiro_Fujinaga/publication/228351016_Automatic_music_classification_and_the_importance_of_instrument_identification/links/09e4150a2948d49dc2000000.pdf
[7] Fala, J. M., Keshap, A. K., Doerning, M. S., & Barbeau, J. T. (1991). U.S. Patent No. 4,991,488. Washington, DC: U.S. Patent and Trademark Office.
[8] Dieleman, S., Brakel, P., & Schrauwen, B. (2011). Audio-based Music Classification with a Pretrained Convolutional Network. In 12th International Society for Music Information Retrieval Conference (ISMIR-2011), 669-674. University of Miami. Retrieved from https://biblio.ugent.be/publication/1989534/file/6741798.pdf