Overview of Automatic Indian Music Information
Recognition, Classification and Retrieval Systems
Trisiladevi C. Nagavi Nagappa U. Bhajantri
Department of Computer Science and Engineering Department of Computer Science and Engineering
S J College of Engineering Government Engineering College
Mysore, Karnataka, India Chamarajanagar, Karnataka, India
tnagavi@yahoo.com bhajan3nu@gmail.com
Abstract- In recent time conventional ways of listening to
music, and methods for discovering music, such as radio
broadcasts and record stores, are being replaced by
personalized ways to hear and learn about music. There is
abundant research and experimentation done over Western
music. However, moderate amount of work noticed over
Indian music and its related fields such as computational
musicology and artificial intelligence to the realm of Indian
music [1]. On the other hand, Indian music has adequate
history and slowly it is becoming an international
phenomenon. This has opened a wide opportunity for an
Indian music discovery system which can suggest music based
either on known artists or on simple descriptive terms. Hence
researchers from various domains such as music processing,
computer engineering and artificial intelligence etc. are
contributing more towards Indian music processing.
Eventually Indian music is broadly classified into South Indian
Carnatic music and North Indian Hindustani music. Both the
systems of music are rich in their own style and Carnatic music
is much more complex in the way the notes are arranged and
rendered [11]. Apart from these two broader categories, there
are various types of Indian music. They are folk, tribal,
bhajans or devotional, bhangra, Indi-pop, film songs, fusion
and ghazals.
This paper presents an overview of previous works on
automatic Indian music information recognition, classification
and retrieval. Furthermore the comparative study of the
recognition, classification and retrieval techniques effectiveness
based on various factors is also presented.
Key words: Hidden Markov Models, Music Information
Retrieval, Artificial Neural Networks, Pitch Class Distribution
and Bayesian Decision Rule.
I. INTRODUCTION
Content based online music enabling systems are being
developed and they are based on melody features. Melody can be
defined as a series of notes sounding in succession. Also it is a
sequence of pitches and durations. The basic elements of a melody
are pitch, duration, dynamics and timbre, where pitch is perceived
fundamental frequency, and duration is particular time interval.
Further dynamics refers to softness or loudness of a sound and
timbre which is also called as tone color is the quality or sound of a
voice or instrument.
Coming to online music access, we can say Music Information
Retrieval (MIR) is the interdisciplinary field of retrieving
information from music. In many publications [5, 6, 7, 8, 14] of
content-based MIR, musical concepts such as melody or harmony,
similarity measures between melodies, and finding occurrences
and variations of a musical structure within a melody are used.
Eventually, due to the encouragement of ardent music lover,
uncountable numbers of people have contributed for the growth of
Indian music in a larger extent. The main concentration of Indian
music researchers is automatic classification of ragas which has
lead to a rich debate on the essential characteristics of raga and
the features that make two ragas similar or dissimilar [3]. Raga is
defined as a characteristic arrangement or progression of notes
whose full potential and complexity can only be realized in
exposition. It is important to note here that no two performances of
the same raga, even two performances by the same artist, will be
identical [2].
The majority of research contributions fall into two broader
categories in order to use Indian music information from MIR. The
first category models music as linear strings and the second one
models music as sets of two-dimensional geometric objects. After
elucidation of melody, MIR and Indian music, we investigate some
of research contributions dealing with identification of raga, swara
etc. from a collection of music recorded as Musical Instrument
Digital Interface (MIDI), MP3 and audio files.
Our paper has attempted to portray the overview of automatic
Indian music information systems using different feature extraction
algorithms and classifiers. Subsequently in section II effectiveness
of the recognition, classification and retrieval techniques are being
compared based on various factors are comprehended. Section III
is a discussion and result analysis of previous works. Conclusion
of the paper is presented in section IV.
II. AUTOMATIC INDIAN MUSIC INFORMATION
SYSTEMS
The specialization of MIR for Indian Music is considered in
comprehensive publications covered by research community.
Majority [4, 6, 7, 12, 14] of content-based MIR are engaged in
intelligent, automated processing of music. Many of them have
addressed some of the recent developments in content-based
analysis and retrieval of music, they paid particular attention to the
methods by which important information about music signals and
symbols can be automatically extracted and processed for MIR
systems. Automatic music information system is an ideal example
of multidisciplinary research. Nowadays, there are a large number
of research works focused on the automatic music information
recognition, classification and retrieval by methods of acoustic
analysis, feature extraction, neural network and statistics. This
section presents an overview of previous works found in the
literature which concentrates on how the automatic music
2011 International Conference on Recent Trends in Information Systems
978-1-4577-0792-6/11/$26.00 ©2011 IEEE
111