基于眼电的字符输入系统研究  

Research of Speller System Based on EOG

作　　者：郑敏敏[1] 高小榕[1] 

机构地区：[1]清华大学医学院生物医学工程系,北京100084

出　　处：《中国生物医学工程学报》2012年第6期801-806,共6页Chinese Journal of Biomedical Engineering

基　　金：国家自然科学基金重点项目(91120007);国家高技术研究发展计划(863计划)(2012AA011601)

摘　　要：为有肢体运动障碍的群体提供一种不须依靠肢体进行信息交流的人机交互系统,使用者可以通过控制眼球视线方向实现与外界进行信息交流。设计了一种注视目标在屏幕上从中心往不同方向跳动的人机交互界面,利用水平和垂直眼电信号检测眼睛注视这些跳动目标时的眼动。算法步骤是首先采用数学形态学法去除眼电中的肌电和眨眼干扰,然后采用主成分分析法提取出不同眼球视线方向的特征向量,对眼电信号进行空间滤波,依据最大幅度所在方向和极性值判断眼球跳视方向,从而确定受试注视目标。结果表明,受试能够通过注视跳动目标,根据自己的意愿进行字符输入的操作。8个受试的结果表明系统的平均信息传输率为25.7 b/min。结合数学形态学和主成分分析的方法能够较好地提取出眼球跳视方向,并且可以转化为控制外界设备的命令。The aim of this work is to provide a human-computer interaction system for persons with some physical disabilities.It allows subjects to control devices by controlling the gaze of direction the eye.A human-computer interface which displays targets repeating moving between the center and eight around directions was designed.The system could detecte the directions of the eyes movement by analyzing the horizontal and vertical EOG when the eyes follow on the certain moving targets.First,the EMG and blink interference were removed from the horizontal and vertical EOG using the mathematical morphology processing.Next,the feature vectors of different eye motion directions were extracted by principal component analysis（PCA）.At last,the smoothed EOG signals were projected onto the feature vectors to detect the directions of the eye movement which were corresponding to a moving target.Subjects could operate the speller system in accordance with one＇s wish and input a character.A total of 8 subjects＇ results were analyzed and the average information rate（ITR） was 25.7 bits/min.The combination of mathematical morphology and principal component analysis methods was feasible,effective and could extract the information about the directions of the eyes movement.Moreover,the information could be transformed into commands to control external devices effectively.

关 键 词：人机交互系统 数学形态学 主成分分析 信息传输率 

分 类 号：R318[医药卫生—生物医学工程]