| The gaze estimation technique is a technique to obtain the current direction of the user’s gaze by using different detection methods.Gaze estimation plays an increasingly important role in our lives.How to improve the accuracy of gaze estimation when the head movement changes is a challenge in the field of gaze estimation at present.This thesis presents a gaze estimation model using eye movement features combined with head movement changes.The thesis focuses on the following:1)Extraction of eye movement featuresIn this thesis,the features of eye movement mainly include the center of the pupil,the canthus point,and the amount of eye movement formed by them.Firstly,the Haar feature and Adaboost algorithm are used for face detection.Then face feature points were detected,and eye images were located and obtained by the ASM algorithm.Finally,dual Haar-like feature extraction,edge filtering,and morphological pixel model were used to accurately locate the pupil center.2)Feature extraction of head motion changesIn this thesis,the feature extraction data of head motion are composed of six degrees of freedom data,including roll angle,pitch angle,yaw angle,and translation along X,Y,and Z directions.Using the EPn P algorithm combined with face feature points,the features are mapped into two-dimensional space through three-dimensional transformation,and the parameters of the model are changed to calculate the attitude data of the current head space position.3)The gaze estimation model of Head-eye movement data fusionBased on the research environment of the laboratory,static and dynamic experiments were designed to evaluate the effectiveness of the proposed gaze estimation model.The dynamic experiment included single head rotation movement and mixed head movement gaze data acquisition experiment with three kinds of rotational movements.The camera is used to collect the image data,and the coordinate of the target points corresponding to the images was recorded for gaze estimation.This thesis discusses the coordination between eye movement and head movement and combines the characteristics of head movement changes and eye movement changes.The polynomial mapping model was used to calculate the initial gaze,and then the deviation effect of head motion changes on the gaze was studied to obtain the compensated gaze.To explore the influence of head motion changes on gaze estimation,the gaze estimation with the three head rotation motion modes was analyzed respectively.Then,the gaze estimation under the composite head motion changes was analyzed.In this thesis,the average error of gaze estimation is about 2.32° by using the compensation model of gaze estimation under the state of head motion change. |
