Study of Machine Learning Methods for Voice Biometric Identification in Data Protection Systems

This research investigates the effectiveness of machine learning methods in biometric voice identification, focusing on their application in data protection systems. Biometric voice identification has emerged as a promising approach due to its ability to analyze unique voice characteristics, enhancing authentication reliability and resistance against spoofing attacks. However, factors such as emotional state and physical health can impact voice consistency, posing challenges to accurate identification. The study involves a comparative analysis of various classifiers, including Gaussian Naive Bayes, support vector machine, and k-nearest neighbors, utilizing a prepared voice biometric dataset. The findings reveal that the Gaussian Naive Bayes classifier outperforms the others in terms of accuracy, achieving a mean accuracy of approximately 88.4%. Despite this, the research acknowledges the necessity for further refinement in preprocessing techniques and dataset enhancements to address existing identification accuracy limitations. Overall, this work emphasizes the potential of machine learning to bolster voice biometric systems while identifying areas for ongoing development..