The Research Of Eye Movement Detection Method Based On The Combination Of EOG And Facial Video

The video-based detection and tracking techniques are commonly-used in eye movement research.Although video-based methods are more convenient to use,it is very sensitive to the change of ambient light.Especially under the conditions of low illumination,the video frames containing informative eye movement cannot be effectively detected,which will lead to significant performance degradation of eye movement detection and classification algorithms.The electrooculogram(EOG),a type of bioelectric signal,can provide valuable information with respect to different eye movement patterns.As compared to other bioelectrical signals,EOG signals are easy to be detected due to the high amplitude.In addition,EOG signal is less affected by the ambient light.Taking into consideration of the advantages of EOG,a novel eye movement detection method involving EOG and facial video was proposed in this thesis.Speicifically,EOG signal is employed to locate the accurate time slices related to informative eye movements.On this bais,we extract useful video frames in terms of the results of EOG detection in order to improve the performance of eye movement analysis system.This thesis mainly concerned with two eye movement patterns,the blink and the reading which includes a variety of eye movement types such as saccades,fixation and so on,and carries on the detection and analysis.The main works can be summerized as follows:(1)Designed four kinds of light variation experiment paradigms and three kinds of reading types according to the requirements of algorithm research.Realized a synchronization acquisition system for recording EOG and facial video data.Furthermore,we established a small eye movement database including EOG and facial video.(2)Proposed an algorithm for implementing EOG-based reading activity recognition.Employed the differential algorithm to detect blink based on EOG.The experimental results showed that the eye movement detection algorithm based on EOG has high accuracy,which verified that the EOG is less affected by changes in the light environment.(3)Proposed an algorithm for implementing blink detection.The algorithm is employed to determine the moments that a blink starts and ends,for the purpose of calculating blink frequency and duration.Converted the image of the two-dimensional eye region into a one-dimensional sequence signal,and employed the reading activity recognition algorithm to detect reading based on facial video.The experimental results showed that the eye movement detection algorithm based on face video would be affected by ambient light,therrfore the higher the degree of light change is,the lower the accuracy is,and even cannot be recognized.(4)Proposed a method based on the combination of EOG and facial video for eye movement detection.EOG signal is employed to locate the accurate time slices related to informative eye movements,and we extract useful video frames in terms of the results of EOG detection in order to improve the accuracy of eye movement detection.The experimental results showed that the proposed method can improve the accuracy of eye movement detection.(5)According to the proposed reading activity recognition algorithm,a reading auxiliary system based on EOG is designed and implemented to assist patients with Dyslexia or the aged maintain fluency in reading.And the system can collect the reading EOG of subject in real time through the bioelectric acquisition equipment,and encode the reading EOG by using the algorithm mentioned in the thesis,and then convert the coding results into the control command as well to realize the control of the magnifying glass in the reading auxiliary system.