Evaluating performance proficiency for a chord performance on guitar using fuzzy AHP

摘要

Previous studies have been reported that aim at automatically evaluating the proficiency of musical performances. Many of these studies have used linear evaluation models on which evaluation parameters are assumed as to be independent. However, subjective evaluation is thought of as nonlinear because of mutual relationships among evaluation parameters. Therefore, this study has developed a method to automatically evaluate proficiency using a fuzzy analytic hierarchy process (AHP) for a chord performance on a MIDI-guitar. Steps of the proposed method are (i) calculating scores for evaluation parameters, and (Ⅱ) calculating the λ fuzzy measures. In step (i), evaluation parameters are calculated from a chord performance on a MIDI-guitar. The evaluation parameters used here are “occurrence frequency of missing notes”, “onset-time deviations from a metronome”, “deviation of MIDI-velocities”, and “deviation of performed notes duration”. These parameters are evaluated in accordance with each evaluation criteria. In step (Ⅱ), the relative weights of evaluation parameters are obtained on the basis of the results of a paired comparison method using stimuli sounds. The parameter λ is determined subjectively by a guitarist who points out that some lucks would give a bad evaluation, with blind respect to several good features. The total evaluation proficiency score is calculated by the Choquet integration, using the scores as evaluation parameters along with the obtained λ fuzzy measures. The proposed method has been confirmed to perform at the same level as a person's subjective evaluation when compared to a simple linear evaluation model.