Loading…
NECEC 2017 has ended
View analytic
Wednesday, November 15 • 12:00 - 12:20
Musical Instrument Recognition Based on Vocal Structure Features and Human Hearing Characteristics

Sign up or log in to save this to your schedule and see who's attending!

In this paper, a musical instrument recognition system is proposed. To realize the system, first we obtain the monophonic musical signals played by various kinds of instruments. Regardless of the relatively complicated feature selection process introduced in other work, we only extract features that stand for the vocal structures of instruments (LPCCs) and characteristics of human hearings (MFCCs), as well as their differential coefficients. By combining those features into feature vectors, certain classifiers (k-NN, SVM) have been selected to classify the vectors into different instruments. The results show that feature vectors containing MFCCs perform better than those without MFCCs. Nevertheless, the combination of the LPCCs and MFCCs yields the best results. For the selection of classifiers, although k-NN requires less training time in our system, SVM will be more favorable when the dimension of feature vectors and the number of samples increase.


Wednesday November 15, 2017 12:00 - 12:20
Conception Bay North 180 Portugal Cove Road, St. John's, NL, Canada

Attendees (1)