"Classification of Indian Songs in Context of Complexity Measure"
Gross classification prevailing in Indian songs are like a) Classical b) Semi-classical and c) Light. The categorization is largely from popular perception of music- nothing very rigorous in nature. Classical songs are composed by following grammar of music more rigorously and consequently have frequent changes in frequency. These types of songs are quite difficult to learn and are supposed to be foundation of all sorts of songs. In light music, in contrast, more stress is given to lyrics and is sung more 'smoothly'. Semi-classical songs lie in between. Here, we shall attempt to find if there is any mathematically foundation of this classification. More particularly, the question is whether we can find a measure based on which Indian songs can be classified. We addressed this issue by exploring complexities associated with the songs with the help of nonlinear analysis of the signals that these songs present. This is in line with our previous work on nonlinear analysis of time series. We collected samples of well-known songs of all these categories for analysis ( Das et al., Complexity, 7: 3, 2002). With appropriate processing of the audio clips, we converted them to time series datasets. We applied nonlinear tools to show that classical songs have a larger fractal dimension than light songs while for semi-classical songs, fractal dimensions lie in between the former two. To the best of our knowledge, complex analysis of Indian songs of the nature presented here are being attempted for the first time. Although, these type of analysis have been applied to western classical- in both vocal and instrumental forms. We have reproduced some of the results of those studies. Finally, based on this conclusion we offer an on-line method for classification of Indian songs of unknown category. This will involve the steps that we have described in the paper- but on a more robust scale.