Questioning the Fundamental Problem-Definition of Mridangam Transcription

There have been several attempts to analyze and characterize percussion instruments using computational methods, both in the context of Western (Sandvold et al., 2004; Tindale et al., 2004) and non-western percussion, specifically on automatic transcription of tabla (Chordia, 2005; Gillet & Richard, 2003) and mridangam strokes (Anantapadmanabhan et al., 2013; 2014). Although Anantapadmanabhan et al. (2013) provide greater insight into the mridangam strokes and their relation to the modes of the drumhead, the transcription approach is limited by its dependency on prior knowledge about the specific modes of the instrument. This puts a constraint for the method to be generalized to other instruments or different tonics (Anantapadmanabhan et al., 2014). Another concern is the unavailability of a unique mapping between the acoustic properties of a segmented stroke and its nomenclature in the vocabulary. It is often observed that the same stroke is uttered differently in the konakkol vocalization (the art of performing percussion syllables vocally in Indian art music), based on contextual variations or grammatical impositions. Most notably, even an expert musician is often unable to resolve such ambiguities on isolated presentation of a stroke. In this paper, we attempt to address this problem by proposing a combination of acoustic and semantic approaches for the contextual transcription of mridangam strokes. We address the problem in an analysis-by-synthesis framework. First, a corpus of mridangam compositions is constructed and annotated. The annotations include both syntactic (i.e. an expert musician adhering to the lexicon without a reference to the acoustic properties of the audio) and listening-based (i.e. perceptual classification by a musician having no exposure to the mridangam repertoire). This facilitates modeling the task from both top-down and bottom-up approaches. In this work, we address the problem of mridangam stroke transcription at the intersection of these two approaches. The rest of the paper is structured in terms of description of the methodology, experimental results, and finally, discussion of the obtained insights.