A resilient voice-based speaker identification approach using deep learning
Baha A. Alsaify, H. S. Abu Arja, B. Y. Maayah, M. M. Al-Taweel, Rami Alazrai · IET conference proceedings. · 2022
Speaker recognition technology has achieved tremendous success over the years as it has become more affordable and reliable. Much research have been conducted in the speaker recognition area, but little progress has been made by the researchers. While deep learning surpassed most machine learning techniques, it has became widely used on the topics of speech technology and its new advanced solutions. In this work, we are proposing a resilient approach for speaker identification with middle-eastern accents. experiments have been carried out in the field of speaker recognition based on deep learning methods. The long short-term memory model was used with Mel-frequency cepstral coefficients. The deep neural network model was used with statistical features and Mel-frequency cepstral coefficients. A spectrogram of each sample was generated as an input for the convolutive neural networks model.