Abstract
This paper addresses musical sounds recognition produced by different instrument and focus on classification of instrument tones. Architecture of back-propagation and networks are applied as classifiers. The discrete Fourier transform vectors, mean, and variance extracted from each segment are used as parameters. The Music Instrument Sample Database (UIOWA) is used for this experiment. The number of instrument is 14. We use 16 different structures of neural networks for recognition these instruments and compare the results. Ezzaidi Hassan obtained SR(14)=12/14 by MLP with 60, 120, and 240 units in the middle layer without impacting of training data set. We obtain SR(14)=13/14 using a different way for analyzing the music sounds.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
