iiet
Indian Institute of Technology Kharagpur

sangitbnrimg-01
sangitbnrimg-02
sangitbnrimg-03

1Decoding Ancient Classifications of Indian Ragas


Ragas have been a fundamental construct of Indian (both Hindustani and Carnatic) classical music for thousands of years. In language theoretic parlance, each raga represents a context free language over the sequence of notes used in rendering that raga. Ragas have well defined grammars in addition to its Arohana (notes in the ascent), Avarohana (notes in the descent), Pakar (signature sequence), vadi and vivadiswaras (admissible and inadmissible notes. Different renditions of the same raga may use remarkably different sequences of notes, but both renditions must satisfy the grammar of the raga. The quality of the rendition depends on the choice of note sequences, as well as other factors such as the timbre (or voice quality), tonal variations (such as meends) and more esoteric features that are perceived by the audience but are hard to define formally.
Given a rendition and the grammar of a raga, determining whether the rendition follows the grammar of the raga is a parsing problem. Indian ragas have been traditionally classified in several ways. Some of the common classification systems are:


  • • Time of day: Ragas are classified into dawn, day, dusk, night ragas and sometimes in more detailed periods of the day/night (such as the prahar of the day/night)

  • • Season: Ragas are classified into seasons such as summer, monsoon, autumn, spring and winter

  • • Mood: Some Ragas are known to offer specific moods, such as serenity, sadness, happiness, solitude, etc

A raga may belong to a combination of these classes. Since a raga is defined by its grammar, it is anticipated that the grammar plays a role in the classification. While this may be true, there also exist prominent exceptions. Two ragas having widely different grammars often belong to the same class. This suggests that it is not the grammar alone that is responsible for such classification – there may be features that are hidden in the traditional renditions of these ragas, which are not explicitly defined, but are responsible in associating a raga with the mood of a season. Such hidden features may have been passed on through generations through the tutelage of traditional ways of rendering a raga, without being explicitly elucidated by the teachers.
The objective of this project is to apply machine learning techniques from artificial intelligence to study hidden features in traditional renditions of various ragas, and to find whether some of these features can be used to explain the classifications of the ragas. A more ambitious next step could be to establish the connection of these features with seasons or time of the day, which in turn would require significant validation from learned as well as uninitiated subjects.
In computer science parlance, the broad tasks are as follows:


  • • Extract note sequences from audio files for raga renditions by expert musicians, parse the renditions into ragas
  • • se machine learning techniques to learn feature based classifiers that explain the classifications of the ragas on the various parallel classification schemes (such as season, time of day, mood)
  • • Validate the features using expert opinion and perceptions of uninitiated audience

Principal Investigator (s)
Pallab Dasgupta, Priyadarshi Patnaik
Co-PIs
Joy Sen; K. S. Rao, Damodar Suar, Sourangshu Bhattacharya

 

 

2Audience Response to Indian Classical Music


Sound and its relation to meaning are considered intrinsic in Indian tradition. This subsection of the project examines the validity of this claim by exploring if Indian classical music evokes similar/identical patterns of evocations in relation to time, location, ambience and emotions in listeners irrespective of their exposure or lack of exposure to Indian classical music. This exploration is interdisciplinary and collaborative, taking its methods from the psychology of music, music and emotion studies, and audience response; it complements the “Decoding of scientific classification of ancient ragas” through machine learning techniques by sharing a common pool of ragas which will be examined by both the methods, and in the future holds the potential for identifying the way music can positively modify both emotions and cognitive efficiency.
The proposed project (sub-section: audience response to music) aims to:


• Examine through “audience response” the validity of the claim that Indian classical music evokes similar/identical patterns of evocations in relation to (a) time, (b) location, (c) ambience and (d) emotions in listeners irrespective of their exposure or lack of exposure to Indian classical music.

• It also wishes to examine if (a) time, (b) location, (c) seasons, and (d) emotional states of audience determine the kind of music they prefer and if this relates to the classification of Indian ragas. As an extension, can such music “evoke” different times, seasons, emotions, etc? – This can have therapeutic implications.


Principal Investigator (s)
Priyadarshi Patnaik, Pallab Dasgupta

Co-PIs
Joy Sen, K. S. Rao, Damodar Suar, Sourangshu Bhattacharya