May Thu Myint
At present, digital music collections are commonly accessed through genre, style, mood, release year, and artist. Music classification based on ethnic traditional style is useful for music analysis and has potential applications in retrieval and recommendation systems. The proposed system is designed to classify for Myanmar ethnic music by using sparse representation classification method to define the class label. In this work, the proposed system present the first attempt to classify audio signals automatically according to their ethnic traditional style, which are characterized by timbres features. Ethnic music classification system has three main steps: preprocessing- the audio piece is converted from stereo to mono channel, feature extraction : after preprocessing, the major nine features are extracted such as zcr, centroid, skew, kurtosis, bandwidth, MFCC mean, MFCCstd, MFCC delta-mean, MFCC delta- std. After the feature extraction, the final step is classification in which these extracted features are classified by using SRC classifier. For classification purpose, learning supervised classifier will be employed to determine the ethnic traditional style of a music piece. In the proposed system, it is used the whole song datasets that is in evaluation of individual feature experiment and feature combination experiment. Also, the only five Myanmar ethnic groups are considered such as Kachin, Kayin, Mon, Shan and Rakhine ethnic songs. In whole song experiment, it used 90 songs for training and testing processes and then classification of overall average accuracy of the five ethnic classes from the 114 features receives 65%. Moreover, in whole song experiment, the accuracies of Kachin, Shan , Mon, Kayin, Yakhine ethnic song receive 50%, 81.64%, 51.68%, 71.66%, 70% respectively.
Myanmar ethnic music; sparse representation classifier; preprocessing;,feature extraction;spectral shape feature