Abstract—This paper proposes a novel method for the
identification of swaras and ragas in an Indian classical
harmonium recital. Swaras are musical notes which are
produced by pressing any key of the harmonium (an
instrument analogous to the piano). Ragas are melodic
combinations of swaras which capture the mood and emotions
of the performance. We first deal with the segmentation of the
audio signal by using two different methods for onset detection
viz. use of spectral flux and by fundamental frequency
estimation. Swara identification is proposed by using the pitch
frequency of the swaras as a distinguishing feature. For this
the frequencies associated with each swara are first identified
and the exact mapping of the frequencies in the given signal is
done by using devised database. In case of raga identification,
an extensive database of commonly used swara permutations is
structured and dynamic programming is used for template
matching and hence raga recognition.
Index Terms—Harmonium, onset detection, raga
recognition, swara transcription.
The authors are with the Dept of Electronics and Telecommunication,
College of Engineering, Pune 411005, India (e-mail:
rpendekar@gmail.com, spm.extc@coep.ac.in).
Cite:Rajshri Pendekar, S. P. Mahajan, Rasika Mujumdar, Pranjali Ganoo, "Harmonium Raga Recognition," International Journal of Machine Learning and Computing vol.3, no. 4, pp. 352-356, 2013.