Navid Khozein Ghanad - Islamic Azad university Of Mashhad, Faculty of Engineering, Department Of ComputerMashhad, Iran
Mohammad Hossein Moattar - Islamic Azad university Of Mashhad, Faculty of Engineering, Department Of Computer Mashhad, Iran
Seyed Javad Seyed Mahdavi - Islamic Azad university Of Mashhad, Faculty of Engineering, Department Of ComputerMashhad, Iran

Raising speed in all classification areas particularly in music signal processing is considered a key challenge. In the present paper, a new method is introduced which it is capable of speeding up the classification of musical instruments. It can raise the speed up to 80 times as high. In the suggested method, the music signal is divided into N equal segments of M length. Then the average samples of each segment is calculated. Once put together, these average samples of segments would lead to a new signal. MFCC feature is then extracted from the newly produced signal. Using SVM classification with the RBF kernel, the musical instruments are recognized. This way, a signal is produced which is M times as small as the main signal and manages to speed up classification. According to the findings, since the newly produced signals are M times as small as the initial music signal, the classification speed is raised for 80 times and Classification accuracy is 94% successful.

musical instrument classification, musical instrument recognition, MFCC feature, noise robustness, speeding up classification, SVM classification