Naad-Nartan UGC-CARE Listed ISSN: 2349-4654
Journal of Dance & music Volume : 12, Issue-1, January 2024
Harmony of tradition and technology : Exploring the evolution of
computational methods in Carnatic music
Priyadarshini S Prof. RNS Saileswari
Research Scholar, Department of Music, Dance & Fine Arts, Department of Music, Dance & Fine Arts,
Sri Padmavati Mahila Visvavidyalayam, Tirupati Sri Padmavati Mahila Visvavidyalayam,Tirupati
Abstract
This article explores the progress made in using computational techniques for music. The goal of this study is to uncover
advancements in applied methods that address aspects of Carnatic music, such as recognizing ragas generating compositions
and analyzing performances. By employing these methods, musicians, scholars and those interested can gain insights into
the intricacies of this musical tradition. To study this, an examination of existing data was undertaken. Additionally, specific
computational tools and algorithms developed for this purpose were studied. The findings of this investigation emphasize the
significance of techniques in music by enhancing the creative and research endeavors of music experts and practitioners alike.
Keywords: Carnatic music, Computational methods, raga identification, data mining, machine learning, signal
processing
Research Paper
Introduction ● To explore various techniques used in music
recognition, music generation, and performance
The Tradition
analysis.
With a history that spans centuries, Carnatic music ● To assess the importance of these computational
is deeply rooted in tradition. Its intricate musical techniques in enhancing the creative research
arrangements, patterns, and spiritual essence set it efforts of music scholars and composers.
apart. In the past learning music was primarily, through
tradition with students relying on their ears and years of Hypothesis
training to grasp and perform compositions. While these The fusion of methods such as data mining, machine
methods are invaluable they do have limitations when learning and signal processing has brought about a
it comes to scalability and objective analysis. However transformation in the exploration and utilization of
thanks to advancements there have been innovations in Carnatic music. These methods have the potential to
studying and practicing Carnatic music. The focus of this improve the recognition of ragas, aid in generating
research is to explore how computational methods can compositions and offer analysis thereby leading to a
contribute to aspects of music such as identifying ragas, profound comprehension.
generating compositions, and analyzing performances.
By leveraging these techniques this study aims to Research Methodology
provide insights, for musicians and researchers alike. To address these objectives, a comprehensive review of
Objectives of the study existing data is done. A literature review encompasses
a survey of computational tools constructed and
The main objectives of this study are: algorithms developed explicitly for Carnatic music
● To trace the development of computational analysis. Further, the possibility of implementation of
techniques in Carnatic music. any one of the tools by a layman without a technical
Naad - Nartan Journal of Dance & music Volume 12, Issue-1, January 2024
Email : naadnartanjdm@[Link]
Harmony of tradition and technology... Priyadarshini S & Prof. RNS Saileswari Vol. : 12, Issue-1, January 2024
background is checked. The findings would emphasize techniques. Early composing tools, such as Max
the role of computational advancements in supporting Matthews' Music IV, became available. These techniques
and enhancing the efforts of music specialists and enabled musicians to play in a methodical manner.
musicians. However, because of its resilience, it has mostly been
The Emergence of Computational Methods utilized in experimental and learning environments.
The early 1980s saw the birth of the musical instrument
The use of computational technologies in the music
digital interface (MIDI)[1] standard, which enabled
industry has become more significant in recent years.
digital control of instruments and components. MIDI
The Electronic and Software Industry has contributed
technology allowed for the construction of digital
a lot to the growth of the music industry. The music
repertoires as well as a standardized means of sharing
available in the cassettes, gramophone recordings, and
musical information between computers and electrical
CDs are collected and processed to match the current
music standards. Among the methods used to process devices.
music data are data mining, machine learning, and The employment of computer technology in music
signal processing. They have opened up new lines of increased between 1990 and 2000. It was also
inquiry and education in the world of music. implemented in Carnatic music. During this time,
Examining the research done in this area, three general significant developments in digital technology, music
categories may be identified– software, and computers for music analysis took place.
1. Analysis of Prerecorded Music and Data: This The digitization of audio recordings of Carnatic music
entails looking at already recorded music and data concerts was a major development in this era. This
using specially created algorithms that use filters, required transferring analog tapes to digital media.
to classify, and arrange them in the appropriate Researchers now had more options to deal with big
order. data[2] thanks to digital audio, allowing for more
2. Composition of Music/Electronic Music: sophisticated analysis and experimentation.
This category encompasses the creation and Computational Techniques in Carnatic Music
composition of music within the genre. Analysis
3. Notation Recognition: This section focuses on
identifying and interpreting notations.
Data Collection and Preprocessing
Early Stages of Computational Music Analysis One of the main difficulties in pre-recorded data analysis
is digitizing audio recordings using computational
The majority of musical research was conducted using techniques and building analytical datasets.[3] Numerous
analog technology before the 1970s. Vinyl records and issues that need fresh research strategies arise from the
magnetic tapes were two common analog media used computational analysis of Carnatic music (Koduri et
for audio recordings. When working with audio data, al.[4] Techniques for preprocessing are those that improve
researchers frequently had to manually transcribe and the audio data's quality and eliminate noise. Techniques
annotate them, which was a labor and time-intensive including filtering, normalizing, segmentation, feature
procedure. extraction, and encoding are employed to eliminate
In the 1960s, early researchers in computer science and noise, extract pertinent information, and enhance the
musicology started to investigate the idea of music notation precision of analysis and modeling.
recognition. These endeavors aimed to create mechanisms For the purpose of collecting and classifying the
capable of transforming scanned or photographed sheet data, metadata[5] related to audio recordings and
music into digital representations. When the first image compositions—such as artist names, details on raga and
scanners became available for research institutes, the tala, concert dates, and locations—is crucial. Data may
Massachusetts Institute of Technology began to apply comprise vocal and instrumental performances, depending
optical music identification on printed sheet music in the on the objectives of the study, in order to examine the
late 1960s (Kassler, 250–254). variations in interpretation and style between these two
In the 1960s - 1970s, basic pitch detectors were categories of performances. Occasionally, scholars gather
developed. These systems could identify key sounds in other data modalities, such as written descriptions, video
a single-note audio recording. These early algorithms recordings, and lyrics, in order to enhance their study and
laid the foundation for advanced phonological analysis contextual comprehension of the music.
Naad-Nartan Journal of Dance & music UGC-CARE Listed ISSN: 2349-4654 109
Vol. : 12, Issue-1, January 2024 Harmony of tradition and technology... Priyadarshini S & Prof. RNS Saileswari
Raga Identification and Classification SVM-Based Raga Recognition
Ragas, the melodic frameworks in Carnatic music, are Machine learning has made machine support vectors
central to its identity. Computational methods have been (SVM)[7] useful for raga recognition. By using the
developed to automatically identify and classify ragas retrieved characteristics to train the SVM models, the
from audio recordings. These methods analyze pitch, researchers were able to discriminate between ragas
tonal patterns, and other musical features to determine with encouraging results (Patil & Kulmethe).
the raga being performed. Deep Learning Applications
[The researcher has been able to try the AI tool ChatGPT
Deep learning approaches are used in raga recognition,
for generating a code to analyze the pitch of a sample
specifically Convolutional Neural Networks (CNNs)
music file, which is a sample of her ongoing PhD research
and Recurrent Neural Networks (RNNs). Improved
on Bangalore composers. The singer is [Link]
recognition accuracy resulted from deep learning
Srinath who has sung one of her [Link]
models' successful automated learning of hierarchical
generated a Python script,[6] which when run on the representations of audio input.[8]
sample music file gave the following pitch distribution.
Composition Generation Using AI and Machine
Learning
While Western music might have already progressed
much into this domain, Indian music is still waiting to
be explored. Although some ideas are already gaining
attention like the "Automatic Music Generation of
Indian Classical Music based on Raga"[9] in which
various machine learning techniques are being
employed, including LSTM, RBM, GCA, FSM, and
GAN producing new music that is stylistically and
acoustically similar to the Indian classical music present
Pic 1. Audio sample in the training dataset.
Notation Software
Like other music traditions, Carnatic music also has
developed its musicography, out of which, a notation
system called Sargam is most commonly practiced.
The Sargam notation scheme which enables easy
music notation storage, publishing, and retrieval using
computers is developed. iSargam[10] (Mammen et al.)
is the only music notation encoding system developed
for Indian music notation. More research work has to
happen in this field.
The integration of modern-day technology in
music education and performance
Pic 2. Pitch Estimation From the time the semiconductor industry started, the
Music Information Retrieval electronic industry has also uplifted the music industry.
The first tanpura from Radel Electronics came to light
MIR-Music information Retrieval is one such field in 1979. Devices required for recording, storing, and
in which much research is done regarding Music propagation of music have developed to such an extent
classification, Music source separation and instrument that everything is available at their fingertips. Many
recognition, and Automatic music transcription. versions of the electronic instruments have come up. Also,
Typically, a signal processing system is composed of the programming software required for music has grown
many modules, such as segmentation, feature extraction, multifolds. Various apps are available to play tanpura,
model creation, and decoding. metronome, and table or a mridangam saath for the artist.
110 UGC-CARE Listed ISSN: 2349-4654 Naad-Nartan Journal of Dance & music
Harmony of tradition and technology... Priyadarshini S & Prof. RNS Saileswari Vol. : 12, Issue-1, January 2024
Due to the extensive research going on in this field, limitless, despite ongoing obstacles. Computational
common music players like Amazon Music and Spotify techniques will surely be essential to maintaining and
are using Machine Learning and AI to come up with developing this treasured cultural legacy as we seek to
user-specific playlists. Technique analysis is one way solve the riddles of Carnatic music.
AI is changing performance in Western music. AI
Footnotes
systems may evaluate recordings or live performances
using sophisticated algorithms to provide artists 1. A broad range of electronic musical instruments,
with comprehensive feedback on their dynamics, computers, and associated audio equipment can be
articulation, intonation, and rhythm. Research is also connected to one another for the purpose of performing,
being done on using AI as an accompanist for a live editing, and recording music. MIDI (Musical Instrument
Digital Interface) is a technological standard that
music performance[11] (Dannenberg).
specifies a communication protocol, digital interface,
Challenges and Limitations and electrical connections.
2. Big data refers to data sets that are too large or
When incorporating technical aspects into research on complex to be dealt with by traditional data-processing
Carnatic music, there are challenges and limitations. application software.
● Initially, there can be limitations on the quantity of 3. Dataset is a Collection of data
high-quality audio data available for processing, 4. [Link]
particularly for rare historical recordings. Computational_Approaches_for_the_Understanding_
● Furthermore, the cultural and social nuances of of_Melody_in_Carnatic_Music
Carnatic music may be challenging for computer 5. Metadata is the information providing data and not the
algorithms to fully represent. content itself.
● And last, in the developing subject of human- 6. [Link]
DZshz8WEmsj5M4zPXX80Wir?usp=drive_link
computer interaction in music research,
7. Support vector machines, which examine data for
researchers need to strike a balance between
regression analysis and classification, are supervised
automating music and preserving the human max-margin models with corresponding learning
element. algorithms in machine learning.
Future Directions 8. [Link]
carnatic-music-machine-learning-sridhar-ravikoti
As technology advances, future methods of Computation 9. [Link]
in Carnatic music are promising. Researchers are authors#authors
investigating how to incorporate real-time analytics into 10. [Link]
performance, giving musicians instant feedback during iSargam_music_notation_representation_for_Indian_
rehearsals and live performances. Additionally, cross- Carnatic_music
disciplinary collaborations between computer scientists, 11. [Link]
musicians, and musicologists will foster innovation
in this area. Notation Software exclusive for Indian References
Classical Music and availability of it across devices Kassler, Michael. “Optical Character-Recognition of Printed
would be very helpful for musicians and students. Signal Music: A Review of Two Dissertations.” Perspectives
processing applications exclusively to cater to Indian of New Music, vol. 11, no. 1, 1972, pp. 250–54. JSTOR,
music would add significantly in preserving the old [Link]
music data and sharing it among the common platforms Koduri, Gopala Krishna, et al. “Computational Approaches
of social media would enrich the collection. for the Understanding of Melody in Carnatic Music.”.
Proceedings of the 12th International Society for Music
Conclusion Information Retrieval Conference, ISMIR 2011. 263-
For scholars, performers, and lovers alike, the seamless 268.
fusion of tradition and technology in Carnatic music [Link]/publication/220723348_
analysis offers an intriguing new frontier. Our knowledge Computational_Approaches_for_the_Understanding_
of_Melody_in_Carnatic_Music
of ragas, composition, and performance in this age-
Geetha, T., V. Exploiting Carnatic Music Characteristics for
old tradition has been enhanced by computational
Music Content Identification. 2 Jan. 2012, [Link].
approaches. The potential for more research and
net/10603/25329.
creativity at the nexus of tradition and technology is
Naad-Nartan Journal of Dance & music UGC-CARE Listed ISSN: 2349-4654 111
Vol. : 12, Issue-1, January 2024 Harmony of tradition and technology... Priyadarshini S & Prof. RNS Saileswari
R, Vijayakumar. Automatic Recognition and Classification of Carnatic Ragas. 13 Oct. 2016, [Link]/10603/111461.
Patil, Mangal, and Maheshwari Kulmethe. “RAGA IDENTIFICATION BY USING SVM CLASSIFIER.” ResearchGate, Nov.
2022, [Link]/publication/371540407_RAGA_IDENTIFICATION_BY_USING_SVM_CLASSIFIER.
Ravikoti, Sridhar. Identifying Ragas in Carnatic Music With Machine Learning. 7 Feb. 2021, [Link]/pulse/
identifying-ragas-carnatic-music-machine-learning-sridhar-ravikoti.
S. Adhikary, M. S. M, S. S. K, S. Bhat and K. P. L, "Automatic Music Generation of Indian Classical Music based on Raga,"
2023 IEEE 8th International Conference for Convergence in Technology (I2CT), Lonavla, India, 2023, pp. 1-7, doi:
10.1109/I2CT57861.2023.10126388.
Mammen, Stanly, et al. “iSargam: Music Notation Representation for Indian Carnatic Music.” EURASIP Journal on Audio,
Speech, and Music Processing, vol. 2016, no. 1, 16 Feb. 2016, doi:10.1186/s13636-016-0083-z.
Dannenberg, Roger B., "Artificial Intelligence, Machine Learning, and Music Understanding,"in Proceedings of the Brazilian
Symposium on Computer Music (SBCM2000), Curitiba, Brazil, 2000.
112 UGC-CARE Listed ISSN: 2349-4654 Naad-Nartan Journal of Dance & music
