To improve the accuracy of lie detection and make the lie detection conclusion as the approved evidence in criminal proceedings, a low-cost scheme which uses eye movement as a cue for lie detection is proposed. First, a low-cost eye movement recording system is used to record eye movement signals, a segmented weighting-annular hough transform is applied to track the iris, a gradient integral projection algorithm is used to detect blink, and the width of eye opening is estimated by the difference image. Then, a relational model associated the eye movement features and the lying is built. The results of the eye movement feature detection show that the low-cost eye movement recording system is suitable for lie detection, and the results of lie detection show that the proposed model is effective for lie detection or assistant lie detection.