AI-Generated Voice Recognition with Convolutional Neural Network

In recent years, the rapid advancement of Artificial Intelligence (AI) voice synthesis technologies has raised significant security concerns, as these tools can be misused for fraud, impersonation, and spreading misinformation. The increasing sophistication of voice deepfakes poses a serious threat to societies, such as privacy and communications security, in digital media. The growing challenge demands reliable methods to authenticate voice recordings. In this research, we propose a machine learning based model to detect Turkish AI-generated voice recordings, as there is a lack of research and solutions focused on non-English languages. We introduce a robust model that is capable of accurately classifying voice samples as either AI or human generated. We analyzed the model with many datasets of both human and AI-generated speeches with different qualities. The performance analyses results show that the proposed model recognizes Turkish AI-generated voice with acceptable accuracy for many systems.