Fisher-Jenks Classification Technique for Music Raaga Identification
Anitha K, B D Parameshachari, Kiran · 2023
Raga identification is crucial for systems that automatically retrieve musical data (MIR). Identifying the Raga used to construct the music is one of the ways South Indian classical music (Carnatic music) songs are categorized and organized. This paper seeks to categorize audio input based on the raaga being played. It uses a set ratio between the shadja and all of the swaras in a raaga. The audio noise is eliminated using the Discrete Fourier Transform (DFT) and a few filtering methods. The Fisher-Jenks Natural Breaks Optimization Algorithm is used to cluster the frequencies. The highest frequency in a cluster is then regarded as the dominant frequency, and all dominant frequencies are then normalized with the lowest dominant frequency. When comparing the Euclidean Distance between the resulting ratios and the database, the entry with the lowest distance value is determined to be raaga. Testing accuracy of the suggested preparation methods and models on a dataset of ten audio files provided by Carnatic ragas was 91.98%.