Abstract
We introduced the use of modified DCTNet for audio signals feature extraction. The modified DCTNet is a development of DCTNet, with its center frequencies of filterbanks geometrically spaced. The modified DCTNet is adaptive to different acoustic scales, and it can better capture low frequency acoustic information that is sensitive to human audio perception. We use features extracted by the modified DCTNet and put them to classifiers. Experimental results on bird song classification, music genre classification, and artist identification show that the modified DCTNet and RNN improve classification rate, and achieve state of the art performance.
Original language | English |
---|---|
Article number | 3405 |
Journal | Journal of the Acoustical Society of America |
Volume | 140 |
Issue number | 4 |
DOIs | |
Publication status | Published - 1 Oct 2016 |