Strumming pattern recognition from Ukulele songs

摘要

Nowadays, the ukulele is a very popular musical instrument because it is melodious, easy to carry, and inexpensive. However, self-studying on the ukulele is difficult for beginners because the existing ukulele lessons lack necessary information for beginners, i.e., ukulele strumming patterns. Strumming is using the dominant hand for playing a string instrument while the other hand holds down notes on the fretboard. Normally, one song uses only one strumming pattern that depends on a song rhythm. This research aims to develop an approach for strumming pattern recognition from ukulele songs. The approach uses a decision tree technique to construct a model for predicting three strumming types: strumming up (u), strumming down (d), and noise (n). Then, those strumming types are passed through a process of strumming pattern summarization in order to recognize a strumming pattern of each song. The experiments use 10 ukulele songs for accuracy testing. The results show that the strumming type predictor gets 84.94% of F-measure by average. Nevertheless, the proposed method could correctly recognize strumming patterns from every ukulele song.