基于几何信息的眼部异常运动分析  

Analysis of Abnormal Eyelid Movements Based on Geometric Information

作　　者：李皓弘 金立左[1] Li Haohong

机构地区：[1]东南大学自动化学院,江苏南京210096

出　　处：《工业控制计算机》2021年第6期30-32,36,共4页Industrial Control Computer

摘　　要：提出了一种眼部异常运动分析方法。基于面部运动障碍疾病患者的面部视频,进行人脸对齐并计算得到眼睛相对开度波形这一几何信息。考虑到患者面部表现中眼部外观极其复杂,使用了基于经验模态分解(EMD)的噪声统计特性去噪方法和基于闭眼检测的方法,对眼部相对开度时序波形进行修正。基于此,在处理后的波形上进行采样,并使用支持向量机(SVM)对样本进行眼部异常运动行为的分类,从而构建眼部运动检测器,实验取得了令人满意的结果。在这些工作的基础上,可实现眼部异常运动频率、长闭眼时长占比这些与面部运动障碍疾病强相关特征的提取。实验结果表明,在患者面部视频数据集上,用该方法提取的特征能很好地与人工标注的特征相符合。所述的分析方法可为面部运动障碍疾病的临床诊断与辅助治疗提供有价值的参考。A method for analyzing abnormal eyelid movements is presented in this paper.Based on the facial video of patients with facial movement disorder,face alignment is performed and the geometric information of the relative eye-opening waveform is calculated.Considering that the eyelid appearance in the patient's facial performance is extremely complex,the noise statistical characteristics denoising method based on empirical mode decomposition(EMD)and the method based on eyes closure detector are used to correct the time-series waveform of the relative eye-opening.Based on this,samples are taken on the processed waveforms,and support vector machines(SVM)are used to classify the behavior of abnormal eyelid movements,thereby an eyelid movement detector is constructed.The experiment has achieved satisfactory results.On the basis of these work,the frequency of abnormal eyelid movements and the total proportion of the duration of long-term eyes closure,which are strongly related to facial movement disorders,can be extracted.The experimental results show that on the patient's facial video dataset,the features extracted by the method in this paper are in good agreement with the manually labeled features.The method described in this paper can provide a valuable reference for clinical diagnosis and auxiliary treatment of facial movement disorders.

关 键 词：眼部异常运动 波形修正 眼部运动检测 患者面部视频数据集 特征提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] R770.4[自动化与计算机技术—计算机科学与技术]