Using the multi-layer perceptron for recognizing the music genres

Бесплатный доступ

The article shows the results of recognizing the music genres by using the multi-layer perceptron. It describes a method to extract classifying parameters from audio files with music compositions, shows a list of characteristics used in the article. Multi-layer perceptron structure and its learning algorithm are both described. The learning error shows a dependency from a value of the learning coefficient and a number of hidden neurons. Recognizing accuracy is tested on a set of music compositions related to five different genres GTZAN containing 1000 examples.

Classifying parameters of music compositions, genres recognition, fast fourier transform, multi-layer perceptron, fast descending algorithm, backpropagation algorithm

Короткий адрес: https://sciup.org/148204777

IDR: 148204777

Статья научная