A COMPARATIVE STUDY OF CARNATIC AND HINDUSTANI RAGA SYSTEMS BY NEURAL NETWORK APPROACH

SRIMANI P.K.¹, PARIMALA Y.G.²*
¹R&D Division, B.U., DSI, Bangalore- 560078, Karnataka, India
²City Engineering College, VTU, Bangalore-560078, Karnataka, India.
* Corresponding Author : ygparimala@yahoo.com

Received : 25-10-2012     Accepted : 06-11-2012     Published : 10-11-2012
Volume : 2     Issue : 1       Pages : 35 - 38
Int J Neural Network 2.1 (2012):35-38

Conflict of Interest : None declared

A unique Neural network approach has been used in the present investigations and a comparative study of the raga systems of Carnatic (CCM) and Hindustani classical music (HCM). The paper concerns a detailed study of the melakartha-janya raga system of CCM, Thaat-raaga system of HCM and cognitive studies of the same based on Artificial Neural networks (ANN). For CCM, studies were confined to the 72 melakartha ragas. For HCM 101 ragas were considered. Relative frequencies of notes in the scales were used as inputs. 100% accuracy was obtained for the melakartha system of CCM for several network topologies while highest accuracy was about 80% in case of HCM. Several networks, namely MLP, PCA, GFF, LR, RBF, TLRN were analyzed and consolidate report was generated.

Carnatic classical music, Hindustani classical music, thaats, melakartha ragas, cognition.

Indian Classical music consists of 2 forms- Hindustani and Carnatic music. CCM has its origin in Samaveda [17,18] . HCM is said to have come from Persia and adopted mostly in the Northern parts of India [15,16] . HCM follows a 20-Thaat system [Table-1] while CCM defines a 72-melakartha system [Table-2] .

Swaras are formed by the combination of fundamental frequencies and their overtones and are by themselves pleasing to the ears. 7 notes defined in both the forms are S,R,G,M,P,D,N,. They are classified into Prakruti swaras-without variants- S and P; vikruti swaras- with variants- R, G, M, D, N. CCM defines 3 variants each for R, G, D and N and 2 for M (16 variations but 12 swarasthanas or semitones due to overlap of 4 semitones) [Table-3] .
In HCM, 2 variants are defined for each of the 5 notes, (12 semitones). Semitones bear relative frequency ratio with reference note or adhara shruthi, S.

Ragas are musical expressions which continually change with time. Ragas are inherently melodic and encompass other aspects such as microtonal variations or gamakas. Ragas are formed by different permutations of the notes. Such ragas are innumerable due to the immense possibilities of combinations of swaras. The raga classification and features in the 2 systems are compared in [Table-4] .

A neural network is constructed by highly interconnected processing units (r neurons) which perform simple mathematical operations. A neural network can be trained to perform a particular function by adjusting the values of the connections (weights) between elements, so that a given input results in a specific target output. The network is adjusted, based on a comparison of the output and target, until the network output matches the target [1-4] .

Based on the existing theories in both the forms of music we find that an extensive scientific process of classification method exists in Carnatic music which has already been analyzed by the authors using neural network models and very accurate results have been obtained using various network topologies, for 72 melakartha ragas of CCM and also unique features of theses ragas due to tonic shift [13-15] . The present work is motivated by the interesting features and results obtained in the previous works of the authors [5-15] . In this work ANN approach has been used for the first time on HCM system. A comparative analysis has been done using ANN for ragas of CCM and HCM systems.

The data sets for the present investigations consisted of the ragas defined in the respective systems. In CCM only the melakartha ragas were taken for input data sets. The frequencies of notes and semitones of melakrtha ragas were determined from the relative frequency ratios. Data sets consisted of 72 scales each with 7 attributes pertaining to the ascending of the scale. For HCM, 101 scales were used and normalised for equal number of attributes for each scale. Attributes chosen were 10 each for ascending(aroha) and descending (avaroha) and 20 for their combined analysis.

Various ANN topologies were used for investigations- Multilayer percepteron (MLP), Principle Component analysis (PCA), Radial Basis Function (RBF), Probabilistic Neural Network (PNN), Classification Support Vector Machine (SVM), Generalized Feed forward (GFF), Time-delay (TDNN), Time-lag Recurrent (TLRN), Recurrent (RN). Data sets consisting of the input attributes, and decision outputs were constructed for CCM and HCM systems.
Number of exemplars were 72 for CCM and 101 for HCM. ANNs were built, 70% of input samples were used for training, 10%-cross-validation & 20% -testing. 1/2 hidden layers, on-line/ batch processing, Levenberg-Marquardt (LM) /momentum learning rules were used. No of epochs were 1000. MSE, correlation and accuracy reports and summary of best networks are shown in [Table-1] to [Table-5] , [Table-6] .

The classification approach based on ANN could be applied very effectively and efficiently on CCM, whereas on HCM it was quite difficult since the built-in logic in HCM is not that consistent as evident from present results, for e.g.,
a. Although HCM uses a 12-note system with similar frequencies, number of major full scales that are accepted are only 10 unlike 72 melakarthas of CCM.
b. The ragas of HCM are characterized by several abstract features: i)thaats are never sung whereas melakartha ragas are extensively elaborated ii) there is clear distinction between melakartha and janya ragas in CCM while in HCM exhaustive classification of all ragas into such categories is not found iii) aroha and avaroha are not always sampoorna or having exactly 7 notes iv)more than 2 raagas can have exactly the same scale but probably differ only in features like duration of holding a certain note, meend, vadi swar, auspicious time of singing the raga, rasa produced, as noted earlier.
c. In CCM, the sets trained using the different networks and related parameters were tested to obtain outputs indicating : i) whether a given raga is shuddha madhyama or prathi madhyama mela,- This gave 100% accuracy for all NNs; ii) classification of ragas into their chakras and iii) their location within chakras. gave 100% results for LR-0-B-M, LR-0-B-L, PNN-0-N-N, TLRN-1-B-M, RBF-1-O-M, RBF-1-B-M.
d. For HCM, several investigations were performed : i) purely on ascending scale ii) purely on descending scale iii) on full scale-i.e., both aroha & avaroha. Nu. of attributes were normalized for meaningful comparison [Table-5] [Table-6] show comparative results for various ANN topologies, best performing NNs. i) SVM gives best results of 97% for training, 30% for CV and 40% for testing for only ascending. ii) For descending, GFF is the best NN- giving 64.7% for training, 30%-CV and 35%-testing, iii) For full-scale, MLPPCA-1-B-L gives best accuracies of 81%-training, 40%-CV and 35%-testing.

A comparative study of CCM and HCM raga systems were made applying ANN topologies 72 melas of CCM and 101 ragas belonging to 10 thaats of HCM were taken for analysis. CCM system, being highly scientific gave expected accuracies of 100% with several networks, while best accuracies in HCM was limited to about 80% for training with SVM network. The relatively poorer accuracies in HCM system even after using the full scales are attributed to various aspects of HCM like i) limitations in the number of thaats ii) want of precise formulation in attributing a given raga to a certain That iii) a clear definition of the scales. We conclude that there is immense scope for further research in the area of cognitive analysis using this unique ANN approach in the complex fields of Ind. Cl. music.

[1] Bishop C.M. (1995) Neural Networks for Pattern Recognition, Oxford, Oxford Univ. Press.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[2] Haykin S. (1994) NNs Comp Found., NY, MM.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[3] Judd J.S. (1990) Neural Network Design and the Complexity of Learning MA, The MIT Press.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[4] Srimani P.K. (2001) Nat. Conf. Appl. Engg. & Tech., India.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[5] Srimani P.K. (1998) Music for Excellence, New Delhi, India.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[6] Srimani P.K. (2004) World Conf. Ved. Sci. Vig. India.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[7] Srimani P.K. (1997) Proc. World Conf., India.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[8] Srimani P.K., Parimala Y.G., (2009) Proc. Mahila Vijnana Sammelana-5, Mangalore.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[9] Srivatasa P.K. (1992) Research, development & improvisation of S. Ind. musical instruments-Tamboora, Veena, Mridangam, ghatam, Sangeeta Nrtya Navarathri, Visvakala, Bangalore.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[10] Srivatsa P.K. (2006) Proc., Triadic Nat. Conf.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[11] Srimani P.K., Parimala Y.G. (2012) Int. Conf. on methods & Models in Sc. & Tech. IETAN, Jaipur.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[12] Srimani P.K. and Parimala Y.G. (2012) Int. Conf. on Intelligent Computational Systems, PSRC, Dubai.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[13] Srimani P.K. and Parimala Y.G. (2011) Int. Conf. on Computer, Comm. and Information Sc. and Engg., Switzerland.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[14] Bhatkhande V.N., Sangit Paddati H., I-IV.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[15] Subba Rao B.C. and Cl H. (956) Music, 1-6.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[16] Ramamurthy & Rajaratnam Sangita Deepike,2001.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[17] Govinda Dikshitar “Sangraha Choodamani, 1800.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

	Table 1- 10 THAATS of HCM [15,16]
	Table 2- 72 Melakarthas of CCM Showing The 12 Chakras, Melakartha Number, Name And Scale [17,18]
	Table 3- Comparative Table Showing the Semitones in CCM and HCM [15-18]
	Table 4- Comparative Table Showing Raga Classification Schemes N CCM and HCM
	Table 5- Comparative Results for Various Ann Topologies for CM [13] and HCM Systems
	Table 6- Results of Best Performing Networks for CCM and HCM