A Computational Model for Bharata Natyam Choreography

Report as inappropriate Your report has been received

Description

Vol. 8 No. 7 October 2010 International Journal of Computer Science and Information Security

Shared by: ijcsis

Stats

views:: 225
posted:: 11/2/2010
language:: English
pages:: 3

Public Domain

/ 3 Download or Print

Document Sample

(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 7, October 2010

A Computational Model for Bharata Natyam Choreography
Sangeeta Jadhav, S.S Dempo College of Commerce and Economics, Panaji, Goa India. sangeetajadhav@yahoo.com
Sasikumar, CDAC, Mumbai, India. the.little.sasi@gmail.com

Abstract: Indian Classical dance has been the slowest to adopt technology and although choreography is mainly a domain of
creativity, computers can help a lot to ease this especially for rhythmically oriented intricate footwork, Nritta and the
complementary movements of other limbs of the body. The main objective of this paper is to identify and classify the various
angalakshanas of Bharata Natyam. An angalakshanas in Bharata Natyam refers to classification of elementary body movements
like head, eye, neck, hands and leg movements. The research aims at a well organised classification of the angalakshanas along
with identification of various constraints on their co- occurrences and sequencing. The constraints may be physical, aesthetic or
preferential and would be identified through literature study and machine learning from databases of existing dance repositories
and training videos. The work will be useful to enhance the learning/teaching of Bharata Natyam (and other dance forms) and
also for animated choreography.

1. Introduction - To develop a computational model of Upanga are supposed to move in unison with Anga. Anga
Bharata Natyam covering angalakshanas and talas (a lakshana is a classification of elementary body movements
rhythmic cycle of beats with an ebb and flow of various which are as follows- Head movements–Shirobheda; Neck
types of intonations resounded on a percussive instrument). movements -Greevabheda; Eye movements -Drushtibheda;
Any dance form can be conceptually decomposed into some Leg movements –Padabheda; Hand movements –
constituent base movements involving specific body parts, Hastabheda. Some of them are as follows:
and norms for combining them. The combinations are
constrained by physical constraints, aesthetic constraints,
preferential constraints, etc. Treatises like Natyashastra
codify such constraints and rules in a human understandable
form. This research attempts to model these in a
computational framework, using practical experience and
existing literature. This would help evolving better teaching
programs, better understanding of the dance form, enhanced
ability to compare one dance form to others in India and
abroad, use ICT in composing and designing dance
programmes, and so on.
The basic methodology proposed would be to study and
tabulate the various angalakshanas, relate them to tala – a beat
cycle – to study constraints and preferences, and develop a
machine learning model which can generate/critic
movement sequences. We propose to use artificial
intelligence techniques such as neural networks, constraint
programming, evolutionary programming and heuristic
search in this process.

1.2 Structure of Bharata Natyam: Bharata Natyam is one of
the most ancient of all the Indian Classical dance styles.
Although there is no formal notation available for the dance While performing, a dancer can use one Angalakshana like
form, it is a highly formalized dance as described in ancient only the eye for example for a beat or use all the
texts. This dance is known for its grace, purity, tenderness and Angalakshanas e.g. eye, neck, head, legs, hands
sculpturesque poses. Bharata Natyam is based on the precepts of simultaneously for the same beat. Of course, not all
among others - Bharata's Natyashastra and Nandikeshwara's combinations are physically possible. Some are aesthetically
Abhinayadarpana - to mention the most well known theorists of not recommended. Similarly, the sequence of angalakshanas
the dance. From the moment the danseuse enters and until from one beat to next also has such constraints governing
dancer leaves the stage, dancer has to regulate their movements to them. This is the focus of the research we discuss here.
musical accompaniment in strict accordance with the rules of
the dance form. There are two major aspects to this form of dance: 2. Related work in this area: Although many research papers
Nritta (rhythmic dance movements) and Nritya can be found dedicated to Western dance styles like ballet and
(representational dance). Although there is no clear foxtrot, very few can be found for Indian Classical Dance.
demarcation between the two this research will restrict itself to here seem to be few studies related to creating computerized
Nritta or pure dance movements. Nritta is divided into Chari - choreography for any Indian classical dance. However the
walking movement; Karna - balanced composition of hand following papers and articles are related to Indian classical
and leg movements; Angahara - different combination of dance with emphasis on animation: Soumyadeep Paul, et al,
Karna; Mandala - postures with hand gestures. The body limbs [6] describes in their paper about an off-site coach who uses
are classified as: Anga (main body parts); Pratyanga (minor low band-width joint-motion data and can instruct students
limbs); and Upanga (features of the face). Pratyanga and who are even geographically far away. This application does

231 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 7, October 2010

not require the user to wear any kind of equipment. Sumant (sometimes three or four in case of similar ones) to indicate a
Narayan Pattanaik [7] in his article has described the work position of that part for e.g. code starting with SH indicate
done at the National Centre for Software Technology, now Samyukta Hasta which is double hand movement and GB
CDAC in Mumbai on creating a computational modeling of indicate Greeva Bheda which is a neck movement and so on.
human body and movement for animating Bharata Natyam The various positions of a limb are given as a two character
dance. code e.g. AH-pa indicates pataka from Asamyukta Hasta.
2.1 Dance Notation Thus across time SH-an, SH-kat, SH-aw means double hand
movements moving from one position to another position.
One of the problems involved in creating computational
This enables us to define a dancer’s position at a particular
models for dance is that of notation. In the field of Western
point in time as a vector giving the position code for each
Dances 'Labanotation' a system using symbols to write down
limb. For example in following fig.
the movements of dancers, especially in a ballet has been
developed particularly following the research conducted at the
University of Frankfurt [9] and Ohio State University[9].
Using labanotation several software programs have been
developed such as LabanWriter[15], CALABRAN[13],
LabanPad[10&12], Limelight[11] and LifeForms[14],
However most of these programs are useful mainly either for
notating and archiving dance or for animating dance.
Annemette P Karpen , Bharata Natyam dancer and
choreographer since 1977, presented a paper[16],
“Labanotation for Indian Dance” at the 11th European
conference for Modern South Asian studies, Amsterdam
University, July1990. In this she tried to solve the problem of
notation for Bharata Natyam by using Labanotation a notation
This pose cane ne vectorised as follows:
used for Western dance. However the problem was not
[SB, GB, DB, SH, PB].
completely resolved because of the many small movements of
Across time, each element of the vector changes the code. A
Bharata Natyam; hence she tried to adapt Sutton Shorthand
Notation in which she added some extra signs for this purpose sequence of vector represents a dance movement in time
although this was not completely systematized. duration. This structure will be used for further analysis and
Thus Labanotation is not adequate for Bharata Natyam and modeling.
our proposed research focus however is totally different from The next concern addressed was movement feasibility. As
the work mentioned above as it seeks to generate mentioned earlier this could be physical constraint ( sitting and
permutations and combinations of angalakshanas, for a given standing at same time or bending back and front or moving neck
input beat, for which new algorithms may need to be side to side and looking straight at same time is not possible) ,
developed. The idea of creating a new notation is not in the dance theory base( hip and chest movements are not allowed ),
purview of this research at present. preferential etc.
A physical constraint within a vector has been identified. The
3. Proposed Computational Model: The overall objective,
as mentioned earlier, is to build a computational model of a restriction may arise more often not as a general restriction,
dancer’s body limb movements across time in the Nritta but as applicable in certain positions e.g. if neck is in position
context. We started with a study and analysis to identify the 3 at time t1 then in (t1+1) position 1 is not feasible.
various distinct movements possible in Bharata Natyam of Identifying such constraints along with the applicable context
various body parts such as eye, head, neck, hand, leg, etc. We can be a hard task given the exponential number of
have now identified a collection including: combinations possible. Not that while we expect most
• 9 Head movements(Shirobheda); constraints to be either at a particular time point or between
• 1 Neck movement (Greevabheda); two successive time points, it is possible to have constraints
• 8 Eye movements(Drishtibheda); across more than two time units also. We do not consider this
• 39 Leg movements(Padabheda); at present.
• Hand movements(Hastabheda): We propose to collect many instances of possible dance
14 single hand(asamyukta) position vectors (henceforth dp vectors), have them labeled
3 double hand (samyukta) by human dance expert and use them to generate the
Typically at any point in time, a dancer can choose for each
knowledge base of such constraints. We have completed a
body part one of these possibilities. Where there is no explicit
movement, the default position can be considered. Based on simple vector generator and further refinement and labeling
experience, observation and literature we have identified is in process.
movement combinations which are either infeasible or These labeled instances will be used in a machine learning
unusual. This is currently within one time unit (one program to build the model. Inductive learning, evolutionary
count/beat/taal). programming and neural network model are under
For formalizing the model, we have adopted a notation system consideration at present. While a model like inductive
as follows. We use the Sanskrit name for the Angalakshanas learning can produce human understandable knowledge,
as mentioned before; the first two letters of this name is used compared to models like Neural Network, other aspects such
232 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 7, October 2010

as quantity and quality of data and nature of data need to be Dance And Choreography; an Intelligent Nondeterministic
also considered in making a selection. Generator”, ACM, 2009 , 30-34, April
Among the current research challenges are formulations of [3] Tom Calvert, “Animating Dance”, ACM, 2007, May
an appropriate feature vector for learning algorithm, choice [4] Tom Calvert ,Lars Wilke , Rhonda Ryman and Ilene fox,
of learning algorithm, representation of context, “Applications of Computers to Dance”,IEEE,2005,6-12 ,July
identification of parameter outside the dp vector which may [5] James H Bradford and Paulette cote- Laurence,” An
influence the choices, dealing with variability among human Application of Artificial Intelligence to the choreography of
expert judgment etc. Dance”, Computers and the humanities,1995, 29, 233-240
The proposed tools to be studied viz. - trained neural [6] Soumyadeep Paul, Sudipta N. Sinha and Amitabha
networks; genetic algorithms and heuristic searches have Mukherjee, “Virtual Kathakali: Gesture driven
been chosen for the study from among others because of the metamorphosis”, Centre for Robotics, IIT Kanpur,
computational complexity of the problem. The utility of http://www.iitk.ac.in/robotics/project_lists/gesture/kathakali/
artificial neural network models lies in the fact that they can [7] Sumant Narayan Pattanaik ,"A Stylised Model for
be used to infer a function from observations. This is Animating Bharata Natyam, an Indian Classical Dance
particularly useful in applications where the complexity of Form" , Computers in art, design, and animation , John
the data or task makes the design of such a function by hand ansdown, Rae A. Earnshaw (eds.) Springer-verlag New York
impractical. A simple set of angalakshanas or elementary Inc., USA, 1989 .
body movements has following complexity: 9 Head movements;
4 Neck movements; 8 Eye movements; 39 Leg movements; [8] Labanotation, The University Of Frankfurt,
Hand movements: 28 single hand and 23 double hand http://www.gotan.ch/labanotation.html
movements. For choreography of n beats for each of these
[9]http://user.unifrankfurt.de/~griesbec/LABANE.HTML
the permutation and combination will result in (7233408)^n
movement possibilities! The resulting database for any beat [10]http://user.unifrankfurt.de/~griesbec/CHOREOE.HTM
is huge and requires proper techniques for management of L#LabanPad
data too.
[11]http://user.unifrankfurt.de/~griesbec/LIMEE.HTML
4. Current status of research: With a Java program, [12http://user.unifrankfurt.de/~griesbec/LABPDE.HTML
permutation and combination of all the Angalakshanas have [13] http://web.bham.ac.uk/calaban/frame.htm
been generated, without any constraints for a single beat and [14]http://www.lifeforms.com/products/index.htm
the resulting output is more than 2 millions of dp vectors .
These constraints are related to those movements which are [15]Labanwriter, The Ohio State University,
either physically not possible to perform or aesthetically not http=http://dance.osu.edu/3_research_gallery/Labanotation
good movements to look at. For example the leg movements _history.html
have 4 types: walking movements, leaps, circles and static [16]Annemette P Karpen’s website, http://www.klassisk-
postures. All 4 movements cannot be done at the same time indiskdans.dk/13bharata%20natyam_kathakali_odissi_kathak
by the performer and hence this is a physical constraint. An _dans_koebenhavn_klassisk%20indisk%20dans_kas
example of aesthetic constraint could be performing dance trupvej%2098%20a_indisk%20musikforening_chau.htm
by looking at one corner only or looking down; which is
never used for Nritta. Even after considering few such
constraints, the result has been reduced to few thousands and
needs further more methods to reduce the output to a
manageable state. The research idea is at its infancy yet.

5. Conclusion: The tool developed could help in choreography. It
can generate various permutations and combinations of
Angalakshanas in Bharata Natyam by which many mundane tasks of
a choreographer will be reduced. The choreographer shall not have
to think which movements to give for Nritta, every time it occurs in
a music number. This tool can quickly show an entire list of moves
available for a particular beat hence voiding repetition of the
choreographer’s favorite move again and again and thus will create
unusual choreography. It can also be used as a tool in dance
education by introducing students to a variety of new dance
movements and allowing them therefore to explore new ways to
expressively combine those movements.

6. References
[1] Scott de Lahunta,“The Future of Choreographic practice :the
Choreographic Language Agent (CLA)”, 2009, February
[2] Mario Nakazawa and Andrea Paezold-Ruehi , “Dancing ,

233 http://sites.google.com/site/ijcsis/
ISSN 1947-5500