Phoneme-inspired playing technique representation and its alignment method for electric bass database

In plucked string instruments such as electric bass, the attack phase is dominated by non-periodic components resulting from picking noise, while the sustain phase is dominated by periodic components resulting from string vibrations. This phenomenon is analogous to unvoiced consonants and voiced vowels in speech, suggesting the possibility of applying speech phoneme representations to plucked string instrument playing techniques. In this study, we design playing technique labels for an electric bass database by treating the attack phase as consonants and the sustain-to-decay phase as vowels. Furthermore, we employ a phoneme alignment algorithm to obtain the alignment between the playing technique labels and the acoustic signals of the electric bass. To conduct experiments, we construct a electric bass database and apply methods based on hidden Markov models and dynamic time warping. As a result, methods based on dynamic time warping, particularly those incorporating timbre transformations, provided the most accurate alignment.