Combination of Hypercolumn Neural Networks Model with Hidden Markov Model Based Lip-reading for Arabic Language
Alaa Sagheer, Naoyuki Tsuruta, 直之 鶴田, Rin-ichiro Taniguchi, 倫一郎 谷口, Sakashi Maeda, 佐嘉志 前田, サギール アラー, ナオユキ ツルタ, リンイチロウ タニグチ, サカシ マエダ · QIR (Kyushu University Institutional Repository) (Kyushu University) · 2004
In recent years Lip-reading systems have received much attention. Because it plays an important role in human communication with computer, namely it can easily smooth the human-computer communication and lets it closer to human-human communication. The importance of lip-reading becomes clearer when the communication environment is not so suitable for speech perception such as among persons who have permanent hearing problems, or elderly people, or under water activities, or generally in any ill conditioned environment. In this paper, we supposed that a combination of the Hypercolumn Neural Network model with the Hidden Markov Model is used to achieve a lip-reading system for Arabic language. This lip-reading system may work under varying lips positions and sizes. Experiments include different nine Arabic sentences gathered from 8 different Arabian people (Male & Female). Results show the performance of our system based Arabic Language.