Recording the eye movement precisely has developed into a mature techn ology.However,the application field of the commercial eye trackers is very narrow,which is attrib uted to its high cost.Future gaze tracking systems should be low-cost and flexible.Therefore,a low cost and simpl e eye movement recording system is designed.The system consists of two common CMOS cameras and a headset-style holder.One camera is used to capture eye images,and the other one is used to capture the s cene images. The cost of the system is only hundredths of the existing commercial eye tracker's. A segmented weight ing-annular Hough transform is proposed to detect the irises in the eye images captured by our system,and the c enters of the irises are used to estimate the gazing mapping by support vector regression (SVR).Finally,a conveni ent calibration procedure is designed. The experiment results show that the average error of the random positions in sc ene im ages is 1.57°,which demonstrates that the low cost eye tracker is able to be us ed for everyday human-computer interaction.