Abstracts of Papers in the Journal of the Acoustical Society of Japan

In general, clarinet reeds have various levels of ease of playing.A reed with high-performance suitability makes it easy for the player to produce the sound they imagine.Currently, the only way to determine a reed's ease of playing is by the player actually performing with it.In this paper, we aim to assess the possibility of automatically estimating performance suitability based solely on images of reeds without the need for performance.After conducting machine learning with two different imaging methods, the accuracy rates were 0.75 and 0.85, and the F-measures were 0.67 and 0.84.Additionally, the results from Gradient-weighted Class Activation Mapping (Grad-CAM) suggested that the judgment was made on the basis of the colour of the epidermis and the distribution of vascular bundles.These findings indicate the potential for classifying performance suitability using images of reeds by machine learning.