IDENTIFIKASI CIRI MUSIK DENGAN MENGGUNAKAN MEL-FREKUENSI CEPSTRAL COEFFICIENTS (MFCCs)

Rozaq, Ahmad and D., Bima Sena Bayu and Wijayanto, Ardik (2010) IDENTIFIKASI CIRI MUSIK DENGAN MENGGUNAKAN MEL-FREKUENSI CEPSTRAL COEFFICIENTS (MFCCs). EEPIS Final Project.

Preview

PDF - Published Version Download (3705Kb) | Preview

Abstract

Indonesian Art Robot Contest (KRSI) is a new division in the series of events of the Indonesian Robot Contest. In this competition each of robot is required to be able to dance following the music. So that the robot can do the job well in the contest, then needed a system that can recognize the special characteristics of the musical accompaniment. To recognize the special characteristics of the musical accompaniment, in this final project is created system identification of musical characteristic with mel-frequency cepstral coefficients (MFCCs). TMS320C6713 DSK is used as a voice signal processing system. Voice signals are processed with filter bank, then be synthesized to obtain a combined frequency which have been isolated. Signals are processed further, with framing windowing then the MFCCs (FFT, Log, IFFT, Lifter, cepstrum FFT) process. The results of the MFCCs is normalized, so that could be used as input for artificial neural network (ANN), ANN output is used to give a decision in calling a stored memory dance motion on the microcontroller Based on the results of testing, a robot or machine can be run with the sounds of music. Motion of robot based on coefficient patterns of MFCC. There are 12% coefficient patterns that can be used for ANN learning with convergent error. MFCC coefficient pattern, which is not used ANN learning can still be recognizable, it shows a similar with patternt that used for learning. Suitability of dance motion with music is 37%, it shows the suitability of motion is less. Key words: signal, filter bank, frequency, microcontroller, ANN, coefficient, MFCC.

Actions (login required)