Abstract:

Considering a large population of guitarist and a relatively poor selection of guitar scores, there should be a certain demand for systems that automatically arrange scores for other instruments to guitar scores. This paper introduces a framework based on hidden Markov model (HMM) that carries out ``arrangement'' and ``fingering determination'' in a unified way. The framework takes forms and picking patterns as its hidden states and a given piece of music as an observation sequence and carries out fingering determination and arrangement as a decoding problem of HMM. With manually-set HMM parameters reflecting preference of beginner guitarists, the framework generates natural fingerings and arrangements suitable for beginners. Some examples of fingering and arrangement generated by the framework are presented.