Rizky Ardiansyah, Rieke Adriati Wijayanti, Berliana Bastiar
This study is based on language differences when making friends or meeting Germans. This research aims to develop a system to recognize the sound of the headset. Using the Convolution Neural Network (CNN) method that is able to improve spelling and Melfrequency Cepstral Coefficients (MFCC) extraction which functions to classify sounds by deciphering or describing the frequency of sound signals by utilizing Raspberry Pi devices to function as a link between the headset and the system which has the potential to work in real time. It is then translated by DeepL and converted into text to speech to produce translated voices. The results obtained using the CNN method obtained the appropriate results, the algorithm is quite good at processing voice data until it becomes text, this is shown at an accuracy value of 97%. © 2024 IEEE.
Politeknik Negeri Malang, Digital Telecommunication Network Department, Malang, Indonesia