Deep learning has achieved immense universality by outperforming GMM and i-vectors on speaker identification. Neural Network approaches have obtained promising results when fed by raw speech samples directly. Modified ...