Text this: Enhancing Arabic text-to-speech synthesis for emotional expression in visually impaired individuals using the artificial hummingbird and hybrid deep learning model