Enhancing Arabic Speaker Identification through Lip Movement Analysis and Deep Representation Learning
Mohammed Imad Ashour, Abbas Saadi Abbas, Saadaldeen Rashid Ahmed, Noor Shaker, Ali Jasim Ghaffoori, A. Hussain, Neesrin Ali Kurdi, Mohammed Al-Sarem, Jamal Fadhil Tawfeq
Abstract
Speaker identification plays a vital role in various real-life applications, including personal safety systems and human-computer interaction. This study presents a novel approach to Arabic speaker identification by integrating lip movement analysis with deep learning techniques. Using the "ArabCeleb" dataset, we conducted data preprocessing and employed a deep learning architecture to extract features from both audio and visual modalities. Our method achieved an impressive speaker identification accuracy of 98.6%. This research contributes to the advancement of speaker identification technology, particularly in Arabic, and provides valuable insights for multimodal speaker recognition. By incorporating lip movement analysis and deep representation learning, our approach enhances the accuracy and reliability of Arabic speaker identification systems. Further research in this direction holds great potential for improving speaker recognition in diverse linguistic contexts.