Automatic musical genre classification of audio using Hidden Markov Model

The rapid growth in audio processing has given much help in advancing the development of digital music. It encourages the creation of method for the genre classification which is able to optimize the learning process to be done with ease, simple and has a good quality in a song search accuracy. Hence we need a development of the learning process with a variety of methods and better algorithms. This study discusses the genre classification with good quality in the classification accuracy using a frequency content characteristics and classification using Hidden Markov Models. From the testing scenario about the parameters of type and filter order, obtained the best parameters are the Butterworth filter order 5. The best system performance has 80% accuracy from the test of 3 genre songs: pop, rock, and dance, with 80% accuracy from the amount of 40 samples data from each genre, with 10 testing data of each genre, quantization characteristic of 20, and 150 iterations for HMM.