Keywords: Indian Classical Music, Emotions, Raga, Rasa, Tala, Timbre, Feature Extraction
I. INTRODUCTION
Indian classical music …show more content…
A raga can evoke more than one rasa. Here, we analyze this relationship by considering different features of raga that contribute to its mood. Sur (Tonality), Taal (Rhythm) and Laya (Tempo) are the basic elements of NICM, so we will analyze their relationship with low level features of musical signal and with high level emotional labels i.e. Rasa.
II. RELATED WORK
Mapping of mood with north Indian classical music involves two stages: audio feature extraction and multi – label classification into moods [2]. The main features of Indian classical music that contribute to emotions are melody and rhythm. The mood categories of class labels are according to the Navarasa terminology.
1] Feature Extraction:
We consider here the melodic features like pitch and progression pattern or signature motif of the raga. Short sequences of three-four notes also elicit some emotions [3]. In Indian Classical Music, the relationship between swara and the frequency or pitch associated with it is not fixed. It depends on the fundamental frequency of the note called as ‘Sa’, which is called as ‘Tonic Centre’ or ‘Scale’ of the song. Fundamental frequencies of all swaras are related to Shruti with a defined ratio as shown below …show more content…
the swara transcript is called as Note Transcription. For Indian classical music, Note Transcription process itself is a very challenging task. Gaurav Pandey used two heuristics, based on the pitch of the audio sample for Note transcription, The Hill Peak Heuristic and Note Duration Heuristic, Sound onset detection and musical meter estimation was proposed by Kalpuri for Note transcription in Indian Classical Music. Surendra Shetty and K.K. Achary provided an approach to extract the six musical features like note transcription, swara combination and aroha, avaroha from the audio using Multi-Layer Perceptrons and Neural Networks [4].
2] Music Classification:
Byeong-jun Han et.al. provide method for mapping musical features to emotions using classifiers Support Vector Regression, Support Vector Machine and Gaussian Mixture Model in polar co-ordinate system with accuracy above 90% [9]. This technique can be applied to Indian Classical Music and Navrasa model. The task of multi-label mapping of music into emotions was investigated by Konstantinos Trohidis et.al. An evaluation of four multi-label classification algorithms was performed using a musical database of over 500 songs. This work can be extended to evaluate mood of North Indian classical music audio stimuli