SIMULTANEOUS ESTIMATION METHOD OF MUSICAL INSTRUMENTS AND TONE BY USING MFCC

摘要

In this paper, a new simultaneous musical instrument and tone estimation method using MFCC (Mel-Frequency Cepstrum Coefficients) for automatic transcription system is proposed. This method consists of two steps. The first step is a musical instrument estimation part with Mahalanobis measure using the extracted a family of features of input music signals. The features of MFCC are compared with those of database music signals which are prepared in advance. In the second step, the musical tone data is determined with the estimation results by measuring the Euclidean distance compared with those of the identified musical database. The effectiveness of the method was examined with 12 types of musical instruments such as Trumpet by evaluating tone estimation results of input music. The simulation results show the high accuracy of tone estimation of the proposed method compared with the conventional approaches based on FFT.