A Combination of Hyper Column Model with Hidden Markov Model for Japanese Lip-Reading System
Year: 2004
Type: conference-paper
Abstract: In recent years, lip-reading systems have received much attention, since they play an important role in human communication with computer especially for hearing impaired and elderly people. In this pa-per, we introduce a novel Japanese lip-reading system combines Hypercolumn Neural Network model (HCM) with Hidden Markov Model (HMM). In this system, we use HCM to extract the visual speech fea-tures from input image. The extracted features are modeled by Gaussian distributions, which is used in rec-ognition phase using HMM. The proposed lip-reading system can work under varying lip positions and sizes. Our experiments were carried out using multiple sentences of Japanese language. All images were captured in a natural environment without special lighting or lip markers used. Experimental results demon-strate that the proposed system performance is supreme.
Authors: Alaa Sagheer, Naoyuki Tsuruta, 直之 鶴田, Rin-ichiro Taniguchi, 倫一郎 谷口, Sakashi Maeda, 佐嘉志 前田, Seiji Hashimoto, 誠司 橋本, 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, セイジ ハシモト
Institutions: Kyushu University, Fukuoka University
Language: English
Topic: Speech and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2262215540