Identification of Dominant Frequencies in Javanese Vocal Phonemes Using Fast Fourier Transform and Random Forest Classification

The majority of speech recognition research currently uses English as the research base, but the results can also be used for another language, including Javanese speech recognition. Previous research stated that there were differences in frequency between English and Dutch. This shows that the frequency of Javanese can also be different. The difference in frequency allows for a new way of recognizing Javanese Speech. By using a dataset of Javanese vowel phonemes, this research aims to identify the dominant frequencies in Javanese speech using the Fast Fourier Transform data extraction an2d the Random Forest Classifier. The feature importance level data will be tested with a deep neural network to determine the accuracy and speed of the process. Choosing a dominant frequency is expected to make the process more effective and efficient in using computing resources.