PEMODELAN PROBABILISTIK NEURAL NETWORK UNTUK KONVERSI SUARA GITAR KE CORD

Abstract

Almost allmusic genreuse guitaras its instrument. Toproducea harmonicguitarvoice needs guitar chords mastery. However. only few peopleareable todistinguish guitar chords. This paper is addressed 10 develop a computational model la convert guitar voice into appropriate cord. In this research. we use Mei Frequency Cepstrum Coefficient (MFCC) as feature extraction because thistechniqueis oftenusedfor voice processing and good enough in presenting thecharacteristics ofasignal voice. Probabilistic Neural Network (PNN) is implemented to classify the [eature into one out of 24 class es of cord. We record 345 for each card (totally we have 8640 recording data with WAV format). Experimenst are conducted for same number of cepstral coefficients (/3. l6. 39 and 5l). with 100 millisecond as time Fame and 40% overlapping betwecn successive Fame. According to the experiment, the maximum accuracy is Y4.31%j(}r 52 number ofcepstral coefficients.