口语阿拉伯数字视觉识别

The main task of this project is to design and implement an audio visual Arabic digits recognizer. I chose to work on Arabic language as few researches and systems are being implemented on Arabic datasets compared to English and other languages. Digits only will be considered in this project for simplicity and limiting the vocabulary size. Some other limitations and assumptions will be made to simplify the task: • Only one speaker is expected at a time. • Frontal pose of the speaker will be assumed. • Occlusion will not be taken in consideration. • Isolated digits are spoken not in a continuous speech.