<![CDATA[This study examines the development of an automatic speech recognition (ASR) system in Sundanese, which still faces data limitations. Dialect variations and the lack of labeled data are the main challenges in the speech recognition process. The approach used is a Convolutional Neural Network (CNN) with Mel-Frequency Cepstral Coefficients (MFCC) feature extraction. The data used were 100 voice recordings consisting of West Sundanese and South Sundanese dialects. The processing process was carried out through the stages of pre-emphasis, framing, windowing, Fourier transform, Mel filter bank, and Discrete Cosine Transform to obtain voice features. The data was divided into 80% training data and 20% test data. The CNN model was then trained to recognize the voice patterns of each dialect. Based on the test results, the model achieved an accuracy of 70% with a loss value of 0.60. These results indicate that the approach used can be applied to limited data, although its performance can still be improved in further research.]]>
Copyrights © 2026
