Research Of Eye Movement Signal Extraction Algorithm And Analysis Of Saccades

Electrooculogram(EOG) signal is bioelectric signal generated by the movement of eyeballs. It can be detected by electrodes pasted around the skin surfaces of eyes. This thesis researches the eye movement recognition algorithm based on EOG signals. The implement method of human-computer interaction based on simple channel EOG signal was studied. And the classification method of saccades in different angles based on support vector machine(SVM) was also studied.Firstly, an EOG signal acquisition system was designed and implemented. The de-noising and endpoint detection algorithm of EOG signals were researched. In order to retain more of the signal details, the dual-tree complex wavelet transform was used in the threshold de-noising method. A modified endpoint detect method was proposed to improve the accuracy of the detected eye movement termination point.Secondly, this thesis researched the recognition algorithm of different kinds of eye movement based on horizontal EOG signals. By analysis of the differential characteristic of eye movement EOG signal, methods of identifying the voluntary blink and the direction of saccade were proposed. Recognition method of 6 eye movement control signals in horizontal was designed and the average recognition rate is 91.15%. The simulation experiment of using eye movement to control electric car has been designed.Finally, the SVM was used in the classification of saccades in different angles. Using the endpoint detect algorithm to locate the start and end point of an eye movement and then calculating characteristic parameters of it. For the first time, choosing amplitude, duration and the extreme value of amplitude during an eye movement as the input vector of SVM. EOG signals of blink, small saccade, vestibular eye reflex and saccades in different angles were classified. When resolution was 4°,the average recognition ratio reaches 85%.