一种视频眼震数据的计算机自动识别方法

Computer recognition of video nystagmus based on digital feature and histogram

目的 判断一组招飞学员视频眼震信号样本的分布,去除眼震图中由于眨眼、主观扫视等异常眼动引起的严重偏离值和干扰值,建立眼震波的计算机自动识别法.方法 采集200例男性招飞学生的眼震数据,采用直方图和Q-Q概率图鉴别样本的分布是否近似于某种类型的分布;通过分析样本的中位数、极差和四分位极差等数字特征,计算数据的上、下截断点,从而去除样本中的异常值（眨眼、扫视等异常眼动波）.结果 通过对眼震数据样本的分析,发现眼震信号若包含较多的异常眼动波时不服从正态分布,去除眼动波（即去除异常值）后的数据样本基本服从正态分布;通过计算上、下截断点去除异常值的方法能去除大部分的异常眼动波,识别剔除率可达94.9%.结论 该方法提供了一种直观判断眼震波数据分布的方法,并可从另一种角度判断异常眼动干扰波的多少,且去除眼动波的效果较好.

Objective To establish computer recognition of video nystagmus by analyzing the sample distribution of pilot candidates＇ video nystagmus with the processing of filtering the deviation and interference caused by blinking and abnormal subjective eye movement. Methods The distribution types of 200 data samples were identified by comparing histogram and Q-Q probability plot. The deviations （caused by blinking and abnormal subjective eye movement） in samples were filtered by analyzing median, range, quartile range and other digital characteristics, and by calculating the upper and lower cut-off point. Results Analysis showed that the spread of nystagmus data was not a normal distribution when certain abnormal eye movements involved until the outliers were filtered. Most of the eye movement deviations （94.9%） could be filtered by processing calculated the upper and lower cut-off point. Conclusions The processing method provides a direct approach to judge the nystagmus data distribution and it can determine the degree of abnormal eye movement interference. Most of the eye movement caused deviations can be filtered by this method.