Recognition of Musical Instruments by Generalized Min-Max Classifiers

The correct classification of single musical sources is a relevant aspect for the source separation task and the automatic transcription of polyphonic music. In this paper, we deal with a classification problem concerning the recognition of six different musical instruments: violin, clarinet, flute, oboe, saxophone and piano. A satisfactory solution of such a recognition problem depends mainly on both the preprocessing procedure (set of features extracted from row data) and the adopted classification system. As concerns feature extraction, a suitable signal preprocessing based on FFT, QFT (Q-constant frequency transform) and cepstrum coefficients are employed. We adopt min-max neurofuzzy networks as the classification model, both in their classical and generalized version. The synthesis of these classifiers is performed by the adaptive resolution training technique (ARC, PARC and GPARC algorithms), since it assures good performances and an excellent automation degree.