Identification and Authentication System of User Voice in Bahasa Indonesia using MFCC and Support Vector Classifier

Over the last decade, deep learning techniques have gained significant attention and are widely adopted across various domains, such as speech recognition and computer vision. This widespread integration is primarily due to the increased availability of extensive datasets, advancements in hardware performance, and continuous improvements in deep learning software and algorithms. These developments have paved the way for applying deep learning techniques to voice recognition. The remarkable ability of deep learning models to learn complex representations from large datasets enables them to recognize individual voices, often surpassing traditional statistical methods accurately. Nevertheless, implementing voice recognition systems for Bahasa Indonesia has faced difficulties due to the language's distinctive linguistic features. The lack of diverse and representative datasets for Bahasa Indonesia is an area that requires further attention and investment. This research aims to present a novel deep learning-based method for voice recognition, explicitly focusing on the challenges and opportunities within the context of the Bahasa Indonesia language. The paper proposes a robust and accurate voice recognition system for Bahasa Indonesia speakers, leveraging Mel-Frequency Cepstral Coefficients for feature extraction from voice data and a Support Vector Classifier for classification. The accuracy of the data speaker recognition reached 90% for the polynomial kernel and 83.33% for the linear kernel.