Classification of Ragas Using Psychoacoustic Features and Soft Computational Techniques

Abstract

Classification of classical melodic structures by style, composer, genre, period, etc., is a rather complex task. The level of difficulty varies across melodic frameworks. It would be interesting to see how we can impart this ability to a machine. In this work, the problem of music classification is taken into consideration with special emphasis on the Raga classification. The challenges and obstacles in creating an automatic music classification system are acknowledged and studied. A new approach for clustering melodies in audio music collections of both western as well as Indian background and its application to genre classification. A simple yet effective new classification technique Mean Centered Clustering (MCC) is discussed. The proposed technique maximizes the distance between difierent clusters and reduces the spread of data in individual clusters. The use of MCC as a preprocessing technique for conventional classifiers like Artificial Neural Network (ANN) and Support Vector Machine (SVM) is also demonstrated. It is observed that the MCC based classifier outperforms the classifiers based on conventional techniques such as Principal Component Analysis (PCA) and Discrete Cosine Transform (DCT). Subsequently, this dissertation reports an improved pattern matching technique for composer and raga classification using a fuzzy analytical hierarchy process-based approach. The technique makes use of class-specific patterns extracted from a pattern discovery technique known as Structure Induction Algorithm for r superdiagonals and compactness trawler. Further, to represent inexact matches a modi ed matching technique is proposed to assign weights to the exact matching scores in a probabilistic manner. Subsequently, the weighted scores are fuzzi ed to quantify the extent of match. Finally, the fuzzy scores are aggregated and classi ed on the basis of minimum Euclidean distance from an ideal solution in the pattern space. Finally, the problem of classification of music structures by using different distributions is taken into consideration. Different popular probability distributions are taken into consideration for this task. The processing is done in both the time as well as frequency domain. MFCC coeffcients are used as a basis to apply distribution estimation.