QUANTIZED MELODIC CONTOURS IN INDIAN ART MUSIC PERCEPTION: APPLICATION TO TRANSCRIPTION Ranjani, H. G. Deepak Paramashivan Thippur V. Sreenivas Dept of ECE, Dept of Music, Dept of ECE, Indian Institute of Science, University of Alberta, Indian Institute of Science, Bangalore Canada Bangalore
[email protected] [email protected] [email protected] ABSTRACT 300 Ragas¯ in Indian Art Music have a florid dynamism asso- M P G ciated with them. Owing to their inherent structural intri- 250 cacies, the endeavor of mapping melodic contours to mu- sical notation becomes cumbersome. We explore the po- ! tential of mapping, through quantization of melodic con- 200 tours and listening test of synthesized music, to capture the Pitch (Hz) nuances of ragas¯ . We address both Hindustani and Car- 150 natic music forms of Indian Art Music. Two quantization schemes are examined using stochastic models of melodic pitch. We attempt to quantify the salience of raga¯ per- 100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 ception from reconstructed melodic contours. Perception Time (s) experiments verify that much of the raga¯ nuances inclu- ! sive of the gamaka (subtle ornamentation) structures can Figure 1. [Color Online] Contour of a vocal melodic clip be retained by sampling and quantizing critical points of (rendition by Vid. Deepak Paramashivan) in Thodi raga¯ of melodic contours. Further, we show application of this re- Carnatic music form (tonic frequency at 146.83 Hz). The sult to automatically transcribe melody of Indian Art Mu- transcribed notes correspond to ‘MaP aGa’. The presence sic.