基于内唇特征提取的唇语识别  

Features Extraction and Recognition of Lip-movement Based on Inner Lip

作　　者：王晓钟[1] 潘保昌[1] 郑胜林[1] 

机构地区：[1]广东工业大学信息工程学院,广东广州510006

出　　处：《计算机与现代化》2009年第2期5-7,12,共4页Computer and Modernization

基　　金：国家自然科学基金资助项目(60572078)

摘　　要：将现有唇读识别常用的双唇线口型模板简化优化,创建了单唇线即内唇口型模板,并结合内唇的灰度特征,利用相关函数的相似性匹配实现唇语识别。该方法在准确有效提取特征的同时降低了运算复杂度,实验数据为单纯的视觉信息,集合为单个发音时,识别率可达90%。实验证明新的尝试具可行性。A new lip model based on the inside lip is proposed. To the existing double lip lines model based on the inside and outside lips which is widely used in the lip-movement recognition, the new model is simplified and optimized. After extracting the inside lip features of gray, the similarity of correlation function is used to achieve the recognition. The method decreases the operation complications and promotes accuracy in the meantime. The experiment gathers visual information of single pronunciation of several voices, and the recognition can reach 90%. Results show that the new trying is feasible.

关 键 词：唇动 唇读 口型模板 特征提取 匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]