Analisis Variasi Jumlah Fitur MFCC pada LSTM dalam Klasifikasi Suara Aksen Berbahasa Inggris

Abstract

Numerous research have been performed to classify English accents using both classic and contemporary classification methods. In general, past research on voice classification and voice recognition employs the MFCC approach to extract voice features. However, neither the fluctuation in the number of cepstral coefficients on MFCC features nor the merits and disadvantages of the MFCC approach in sound categorization are discussed. The purpose of this study is to examine the variation in the number of MFCC features used in the classification of English accents using the MFCC and LSTM approaches. In this investigation, the classification accuracy of English accent sounds was determined to be 64.96 percent when the MFCC feature value was set to 17.