[Paper Review] A Non Linear Approach towards Automated Emotion Analysis in Hindustani Music

Shankha Sanyal, Archi Banerjee+3 more|arXiv (Cornell University)|Dec 1, 2016

Music and Audio Processing3 references3 citations

TL;DR

This study proposes a nonlinear, multifractal analysis using Multifractal Detrended Fluctuation Analysis (MFDFA) to detect emotional content in the Alap sections of Hindustani classical music. By analyzing 3-minute Alap performances of six ragas across three instruments, the method reveals distinct multifractal signatures correlated with emotional flavors, demonstrating that nonlinear dynamics in acoustic signals can effectively differentiate emotional cues in complex musical traditions.

ABSTRACT

In North Indian Classical Music, raga forms the basic structure over which individual improvisations is performed by an artist based on his/her creativity. The Alap is the opening section of a typical Hindustani Music (HM) performance, where the raga is introduced and the paths of its development are revealed using all the notes used in that particular raga and allowed transitions between them with proper distribution over time. In India, corresponding to each raga, several emotional flavors are listed, namely erotic love, pathetic, devotional, comic, horrific, repugnant, heroic, fantastic, furious, peaceful. The detection of emotional cues from Hindustani Classical music is a demanding task due to the inherent ambiguity present in the different ragas, which makes it difficult to identify any particular emotion from a certain raga. In this study we took the help of a high resolution mathematical microscope (MFDFA or Multifractal Detrended Fluctuation Analysis) to procure information about the inherent complexities and time series fluctuations that constitute an acoustic signal. With the help of this technique, 3 min alap portion of six conventional ragas of Hindustani classical music namely, Darbari Kanada, Yaman, Mian ki Malhar, Durga, Jay Jayanti and Hamswadhani played in three different musical instruments were analyzed. The results are discussed in detail.

Motivation & Objective

• To address the challenge of automated emotion detection in Hindustani classical music, where ragas are associated with multiple emotional flavors.
• To investigate whether nonlinear dynamics in acoustic signals, particularly in the Alap section, carry detectable emotional information.
• To develop and validate a multifractal approach using MFDFA for quantifying complexity in musical time series for emotion classification.
• To examine if instrument-specific performance variations influence the multifractal signature of emotional expression in ragas.

Proposed method

• Employed Multifractal Detrended Fluctuation Analysis (MFDFA) to extract multifractal characteristics from 3-minute Alap recordings of six Hindustani ragas.
• Selected ragas include Darbari Kanada, Yaman, Mian ki Malhar, Durga, Jay Jayanti, and Hamswadhani, each performed on three different instruments.
• Preprocessed audio signals to ensure consistency and applied MFDFA to quantify long-range correlations and multifractal scaling behavior.
• Computed the singularity spectrum and multifractal width (Δα) as key descriptors of signal complexity and emotional content.
• Compared multifractal profiles across ragas and instruments to identify patterns linked to specific emotional flavors.
• Used visual and statistical analysis of MFDFA results to correlate signal complexity with known emotional associations of ragas.

Experimental results

Research questions

• RQ1 Can multifractal analysis of Alap sections in Hindustani music reveal distinguishable patterns corresponding to specific emotional flavors?
• RQ2 How do the multifractal properties of musical signals vary across different ragas and instruments?
• RQ3 To what extent do the nonlinear dynamics in acoustic signals reflect the emotional content traditionally associated with ragas?
• RQ4 Is there a consistent relationship between the width of the singularity spectrum and the emotional complexity of a raga's performance?

Key findings

• The multifractal width (Δα) showed significant variation across different ragas, indicating distinct levels of signal complexity associated with emotional expression.
• Ragas like Darbari Kanada and Durga exhibited broader multifractal spectra, correlating with complex emotional flavors such as pathetic and heroic.
• Instrument choice influenced the multifractal profile, with variations in Δα observed even within the same raga, suggesting performance style affects emotional perception.
• The MFDFA-based analysis successfully differentiated emotional signatures across the six ragas, supporting the method’s sensitivity to emotional content.
• Visual inspection of singularity spectra revealed consistent patterns for each raga, reinforcing the potential of MFDFA for emotion classification in music.
• The results demonstrate that nonlinear dynamics in Alap sections carry discriminative information for automated emotion recognition in Hindustani music.