TWI261785B - Eye ball tracking method - Google Patents

Abstract

The present invention relates to an eye ball tracking method, which includes the following steps: (1) five-point calibration step; (2) calibration parameters configuration step; (3) coordinate determination step; (4) grid offset determination step; (5) coordinate conversion step; and other optional post-processing steps: (6) plural moving points acquisition step; (7) moving direction determination step; and (8) output track calculation step. As such, the present invention is characterized by the advantages and performance of simple calculation, fast conversion speed, calibration upon detecting a deviation of user's wearing position, protection against error caused by natural vibration of eye ball, and sensitivity of fine-tuning system.

Description

1261785 玖, invention description: [Technical field to which the invention pertains] The present invention relates to an eye tracking tracking method, and particularly to a method for using a four-quadrant midpoint coordinate corresponding to a simple conversion formula 5 to achieve fast conversion It has the advantages of simple calculation and fast conversion speed, avoiding the error of natural vibration of the eyeball, and the advantages and functions of the sensitivity of the system that can be fine-tuned. [Prior Art]

The traditional eyeball trajectory tracking method includes the following steps: 10 -, five-point calibration steps:

First, the user's eyes look at a target screen 9 1, which may be a small screen (as shown in the first and second figures), a computer screen, a large projection screen (as shown in the third figure), etc. And an eye image capturing device 9 2 obtains an eye image 9 3 (as shown in the fourth figure) by an oblique angle 0 from the position below the eye. For example, as shown in FIGS. 5A to 5E and 6A to 6E, the user respectively looks at the reference point C 0 , the right reference point C 1 , and the middle reference on a target pupil plane 9 1 . Point C 2, left reference point C 3 and lower reference point C 4 have coordinates (X 〇, Y 〇), (X丨, Y丨), (X 2, Y 2), (X3, Y3) and ( X4, Y4), and obtains an eye point (Ρ., Q.) and a right point (Ρ, Q, respectively) by taking an eye image 9 through an eye image capturing device 92. ), look at the central point (Ρ 2, Q 2), look at the left point (Ρ 3, Q 3), and look at the lower point (P4, Q4). Second, calculate the first and second oblique axis angle and the horizontal axis value and the vertical axis value for a long time 6 1261785 Please refer to the seventh and eighth figures, assuming that the center point coordinates of the pupil are (m, n) (not shown), and The coordinates have been taken as (pm, pn), Bay|J Pm =m-P2 Pn 5 If Van 0, Bay ijL 2 PX ln =Pn+PmXt^^ Where, <〇, 贝1J lm =PnJC0^ 2 10 ln =pn -Pmxtan02

Wherein, the first oblique axis angle Θ = a sin

Qi ~Q\ Second oblique axis angle θ2 two a sin _Qi ~ Q3_λ! (P2 - P3)2 + (Q2 - Qi)1 15 Therefore, the horizontal axis value after conversion / /77 and the vertical axis value can be obtained / /7. Third, according to the oblique axis scaling step: The calculation of the corresponding coordinates (Dm, Dn) can be divided into the following four cases: 1. If, 0 and / / 7 2 0, Bay '1

Ά —handle-p/hQ2-q/X] "X〇)Cn (ft-β2) 20 7 1261785 "(z 丨-z0)c21 I(p2-p])2^(Q2-q])2 D=Yt (ft-a) (m)c22 If 7 <0 and //7 20, then = X, handle -RfHQ2-Q3y :(Z2—Xj)C31 (04 -ft) (d)c32

4. If 7 < 0 and / / ? < 0, then

Dni II X' V(A ~^)2+(02 ~Q^y

Dn-Y^

(^-^〇)C42

And wherein Ch, C12, C21, C22, C31, C32, C41, and c 10 are each a predetermined system constant; thereby, the corresponding coordinates (Dm,) on a target pupil plane 9 1 can be calculated. Dn). However, assuming that the number of images taken is twenty sheets per second, and the conventional overall calculation process is complicated, it is necessary to calculate the square, the opening number, etc., if the interval between the 15 images is shorter (0.0 5 seconds) Or faster, or the device is slower to calculate. It may happen that the conversion speed is slow and exceeds 〇.〇5 seconds. It is too late to calculate the dilemma of the next image, which seriously affects the operation of the system. Secondly, the traditional eyeball trajectory tracking method directly converts the eyeball center position of the eye image 8 1261785 t directly onto the display screen, but when the user's eyes move, there is a so-called natural vibration and tremor. Movement in a non-fixed direction (or "cursor drift"), so the actual moving path is irregular, and if not filtered to exclude the movement caused by these involuntary vibrations, the result is The situation of irregular vibration will be exerted, which affects the handling and stability of the entire eye control system.

For example, if the user looks at the moving path map shown in the thirteenth figure and starts from the left end in sequence, moves to the right end along the upper path, and then returns to the starting point along the lower path. However, since the human eye inevitably has involuntary natural vibrations 10, the data of the eyeball moving point actually taken will be as shown by the solid line in Fig. 14, showing an irregular moving path, especially the fixation point. When P 5 to P6 and P6 to P7, the moving direction deviates greatly, which is the irregular movement caused by typical natural vibration, which will affect the handling and stability of the subsequent whole eye control system. 15 Therefore, it is necessary to develop new technologies to address the above shortcomings.

SUMMARY OF THE INVENTION The main object of the present invention is to provide an eyeball trajectory tracking method, which has a simple calculation formula and a fast conversion speed. A second object of the present invention is to provide an eyeball trajectory tracking method 20 which can be corrected when there is a deviation in the wearing position of the user. Another object of the present invention is to provide an eyeball trajectory tracking method, wherein the gaze point of the deviation is filtered out by using the cumulative number of times of the same moving direction, thereby obtaining a smoother and natural vibration-free eyeball trajectory data, overcoming the person. The problem of natural vibration of the eye. 9 1261785 Still another object of the present invention is to provide an eyeball tracking method capable of adjusting the sensitivity of a system by adjusting a preset threshold value of a cumulative number of times in which the moving direction is the same. The above objects and advantages of the present invention are not difficult to understand from the detailed description of the selected embodiment 5 and the accompanying drawings. The present invention will be described in detail below with reference to the following embodiments:

The invention is an eyeball trajectory tracking method, as shown in the ninth figure, mainly comprising the following steps: · one, five point correction step 1 1 , two, correction system 10 number setting step 1 2, three, quadrant discrimination step 1 3 4, the outlier offset judgment step 1 4, 5, coordinate conversion step 15 5. For the five-point calibration procedure of the first step, please refer to the tenth and eleventh figures, the user separately looks at the correction point (S c X, S cy) and the right correction point (GX 1, G y ) on a display screen. 1), upper correction point (GX 2, G y 2 ), left correction point 15 (GX 3 , G y 3 ), and lower correction point (GX 4, G y 4 ), and captured through an eye image

The device captures the image of the eye and obtains the gaze midpoint (MOx, MOy), the gaze right point (Μ x 1, M y 1 ), the gaze upper point (Μ x 2, M y 2 ), and the gaze left point (Mx3). , My3) and watch the next point (Mx4, My4). The second step of the correction coefficient setting step is as follows: 2 0 sets four sets of correction coefficients (久久, eight 1), (long 2, eight 2), (long 3, /^3), (eight 4, 8 4), in practice can be adjusted or modified according to the experience value. Referring to FIG. 12, the third step is a quadrant discrimination step. When a user's fixation point (Μ X, M y ) is obtained, it can be mapped to a screen coordinate (GX, G y ); The positive and negative values of the viewpoint (Μ X, M y ) and the midpoint of the gaze (Μ 0 X, Μ 0 y ) 10 1261785 are (positive, positive), (negative, positive), (negative, negative), ( Positive and negative) are respectively determined as the first quadrant, the second quadrant, the third quadrant, and the fourth quadrant. The fourth step is a step of determining the offset of the outage: 5 If the number of outbounds per second occurs more than a predetermined number of times, it can be determined that the wearing position of the user is deviated, that is, the fixation point (Μ X, M y ) exceeds The correction range should be corrected and the following formula should be calculated: If Mxs>Mx, Bay 1J Ax = Mx-Mxs, No Bay 1J AX = 0 (indicating no correction required); 10 If Mys> My, Bay 1J Ay = My-Mys, No Bay|J Ay = 0 (indicating no correction); where Mxs and Mys are the boundary points of the quadrant, and the maximum out-of-place position offset Αχ and Ay can be calculated; Coordinate conversion steps: 15 See also Figure 12. If it is the first quadrant, the Bay 1 screen coordinates (GX, G y ) are:

Gx=Scx+Ix^PxX{Mx+Ax-Sx,)

Gy^Scy^Iy^Py,{My^Ay-Sy,) where the coordinates of the first quadrant are 2 0

〇_ 〇, ~ ^cx e _ CI ^] 2 ^cv ^χ] ~ ^cx H 2 , Λ.ν1 -~Η 2 and wherein the first quadrant midpoint coordinates of the eyeball position are 11 1261785 , Μχ'—Μ,, 2 Μ If it is the second quadrant, the screen coordinates (G χ, G y ) are

Gx^Scx+Ix2+px2{M^Ax-Sx2)

Gv-Scy^Iv2^PY2{My+Ay-Sv2) wherein the coordinates of the second quadrant are

And wherein the midpoint of the second quadrant where the eyeball position is located is

Mox-Mx 2 Μ 〇y

Myl~M0 2 10 If it is the third quadrant, the screen coordinates (Gx, Gy) are Gx=Scx^Ix^Px, {Mx^Ax-Sx,) Gy^Scy+Iy^Py3{My+Ay-Sy3) Wherein, the coordinates of the third quadrant are

,S cy ~ Gy4 ~2

And wherein the third quadrant midpoint coordinate 15 of the eyeball position is 'M,, λ-3 ',) '3 Μ ν'4 〇y, if it is the fourth quadrant, the screen coordinates (G x, G y ) for

Gx-Scx+Ix^PxA(Mx+Ax-SxA)

Gy = Scy + /y4 + βν4 (My + Ay- Sv4) 12 1261785 wherein the coordinates of the fourth quadrant are C _ Ο I ^.rl ~ ^cx Ο _ C , ^ L'y G '.4 10 0 Μ - 10 0 2.2 and wherein the coordinates of the fourth quadrant of the eyeball position are Μ, Μ.χ'-Μα 2 Μ, ' ν4 2 5 , thereby calculating the screen coordinates (Gx , Gy).

In order to explain the actual operation process of the present invention in detail, a conversion at a certain point of the first quadrant is taken as an example, if five correction points are provided on the target screen of an 8 Ο Ο X 6 0 0 and the center is The point is determined as the origin (0,0), where the middle correction point (Sex, Scy) = (0,0), the right correction point (Gxl, Gyl) = (400,0), and the 10 upper correction point (GX 2 , G y 2 ) = ( 0,3 0 0 ). By taking only the midpoint position, it is easy to find the coordinates of the first quadrant area (2 0 0 , 1 5 0 ).

After the first step of the present invention, the midpoint of the gaze is obtained (ΜΟχ, M0y)-(3 8 0, 2 9 0) > gaze to the right point (Mxl, M y 1 ) = ( 4 8 0, 3 0 0 ) > Look at the upper point (Mx2, My2) = (380, 340). 15 The second step of the correction coefficient setting step is as follows: Set four sets of correction coefficients (々rl, gossip), (long 2, eight 2), (/?λ3, /?), 3), (Ad ), where ^, ;^)^0·98, 0.95). The third step is the quadrant discrimination step. When the user's fixation point (Mx, My) = (450, 320), it can be mapped to a screen coordinate (Gx, Gy); 2 0 by the fixation point (Mx, My) = (4 5 0,3 2 0 ) and the positive and negative values after the subtraction of the midpoint of the gaze (ΜΟχ , Μ 0 y ) = ( 3 8 0, 2 9 0 ) are both (positive and positive) and are judged as 13 1261785 First quadrant. The fourth step is the step of judging the offset: since the gaze point (Mx, My) = ( 4 5 0, 3 2 0 ), the correction range is not exceeded (both less than Mxs and Mys), ie Αχ = 0 (representing No correction is required) and Ay=0 (5 means no correction is required); the fifth step is coordinate conversion step: See also the twelfth figure, because it is the first quadrant, where the coordinates of the first quadrant are &;丨=+ G^~---^=4 00 + (400-0)/2 = 600 10 15

S + less 2 _,..= 3 00 + (300-0)/2 = 450 ' 2 and wherein the first quadrant midpoint coordinates of the eyeball position are

7vl =Mox + ^^-^= 3 8 0 + (4 8 0 - 3 8 0 )/2 = 4 3 0 Μ , - Μ

Iyx = M〇y + 290 + (340-290)/2 = 315 The screen coordinates (Gx, Gy) are Gx-Scx^Ix^PxA{Mx^Ax-Sx,) =400+430+0.98*(480 +0-600) = 7 12

Gy-Scy^Iy^Py]{My^Ay-SvX) 14 1261785 =300+315+0.95*(320+0-450) = 49 1 Therefore, the corresponding screen coordinates (G x, G y can be quickly converted) ) For (7 1 2, 4 9 1 ), the calculation process is simple and fast. 5 Of course, in order to make the eyeball tracking method of the present invention more accurate, in particular, the error of natural vibration of the eyeball can be avoided, as shown in the ninth figure, the following subsequent processing steps can be added: 6. Obtaining a plurality of moving points Steps 16 and 7. Determine the moving direction Step 1 7. Eight, output execution calculation step 18

10 Referring to Figures 13 and 14, the details of the steps of the present invention are as follows: 1. Step of obtaining a plurality of moving points: The user's eyes move a predetermined trajectory and sample at different times (t) to obtain a plurality of moving points. (x(t), y(t)), and two adjacent moving point shapes 1 5 into a moving vector 'as shown in the following formula'

Second, determine the direction of movement steps: Also, ||卩(1) Bu V(x(1))2 + 〇(〇) When v(t-n) I Deng-/7)1

Ap < 20 Using the above two formulas to determine the direction of movement, wherein the user-defined error value, as long as the above formula is satisfied, we call the same moving direction, and define η as the cumulative number of moving directions; 15 1261785 III Output trace calculation step: When the cumulative number of times η of the same moving direction is less than, equal to or greater than a preset threshold k, the output track tile is calculated by the following formula: lv(〇if η < k if n&gt ; k 5 In the above formula, m is a predetermined correction coefficient, whereby the eye movement output trajectory 圮(1) of the natural vibration of the eye can be avoided.

Here we actually enumerate an example of an eye movement obstruction: First we assume that the eye observes the thirteenth figure in a clockwise direction. After that, the solid line part of the fourteenth figure represents the actual movement of the eyeball, and the dotted line part of the fourteenth figure is calculated by the algorithm of the present invention, and the obtained eye movement trajectory is analyzed. Step by step

For convenience of explanation, we assume that the parameter condition in the figure is the correction coefficient m = 1, the preset threshold k = 2, that is, when the eyeball movement direction is accumulated twice, the 15 system determines that the eyeball is moving in the same direction in the past, in fact In the case, the appropriate k value must be obtained through multiple experiments. Of course, by changing the size of the preset threshold k, the sensitivity of the system can be fine-tuned. The eyesight is starting from the A area to view this picture, passing through the B area and the C area, and returning to the A area. It is assumed that in the B zone and the C zone, an involuntary vibration occurs in each of the target balls, and thus the measured position has a different direction and an unfixed direction of the phenomenon of 16 1261785. Where R: represents the measured amount of movement of the i-th time (input) t: represents the i-th cursor movement amount (output) △ household: represents the allowable error offset η: represents the same cumulative number of movement directions

When the fixation point moves from Ρ 0 to Ρ 1, L = (Vly〇) represents the first output result, in which the cumulative number of movements after the first movement is the same. 10 When the fixation point is moved from P 1 to P 2 , =巧=>" The second movement direction of the second and second movements is the same as the movement direction of the first time, so the input and output results are the same and η is increased by 1, and n = 2 at this time, confirming the movement direction of the target ball.

When the fixation point is moved from P2 to P3, $OFF%=^ = 〇 The third movement direction is different from the second movement direction, so η is reset to 0, that is, the cumulative number of times is zero. When the fixation point is moved from Ρ3 to Ρ4, the direction of the third and fourth movements is within the error tolerance range, so the input and output results are the same and η is increased by 1, and η = 1. When the fixation point is moved from Ρ 4 to Ρ 5, ν4 - Δρ < ν5 < ν4 + => νο5 = ν5 2 0 The direction of the four or five movements is within the error tolerance range, so the input and output results are the same and η Increase by 1, at this time η = 2, confirm the direction of eye movement. 17 1261785 When the fixation point is moved from P 5 to P 6 ′ K矣v5 /7 = 0 The movement direction of the six times is different from the movement direction of the fifth time, so η is reset to 0. When the fixation point is moved from Ρ6 to Ρ7, V7 is closed to m = 0, the seventh movement direction is different from the sixth movement direction, so η is also set to zero. When the fixation point is moved from Ρ7 to Ρ8, % off 6=>" = 〇 The eighth movement direction is different from the seventh movement direction, so η is still 0.

When the fixation point is moved from Ρ 8 to Ρ 9,% — Δρ <孓< intersection 8 + Δρ => t = intersection 9 八, the direction of the nine movements is within the error tolerance range, so the input and output results are the same and Increase by 1, where η = 1. 10 When the gaze point moves from Ρ 9 to Ρ 1 0, pay 9 - △ /? < 丨 S 0 S intersection 9 + Δρ => < 丨 0 two intersections 10 IX, the direction of the twelfth movement is allowed in the error Within the range, the input and output results are the same and η is increased by 1, and η = 2 at this time, confirming the direction of eye movement. The rest are analogies, so I won't go into details here.

Therefore, as can be seen from Fig. 14, from the gaze point Ρ 0 to Ρ 1 0, the actual output trajectory of the present invention is Ρ 0 to Ρ 2, then Ρ 2 to Ρ 5, and then Ρ 5 to Ρ 1 0 It is obvious that this output result (shown by the dotted line in Fig. 14) is not only very close to the corresponding moving path shown in Fig. 13, but also the irregular offset point Ρ6 caused by the involuntary natural vibration is filtered out. . If you observe the entire movement trajectory, according to the above method, as shown in Figure XIV, the algorithm continues to track the trajectory of eye movement, starting from the squatting area, passing through the B area and then returning to the A area, all the analyzed The output vectors are connected together to obtain the dashed line in the fourteenth figure, that is, the trajectory of the eye movement obtained by the algorithm analysis after 18 1261785, in which there is a natural vibration in the B area and the C area due to involuntary human eyes. The resulting non-fixed direction of movement, under the judgment of this algorithm, can be distinguished not by the direction in which the eyeball actually moves. 5 In summary, the advantages and effects of the present invention can be summarized as follows: [1] The formula is simple and the conversion speed is fast. Since the calculation method adopted by the present invention is simple, and there is no cumbersome square, root number and lengthy calculation, the actual calculation time can be shortened.

[2] It can be corrected when there is a deviation in the user's wearing position. Since 10 users occasionally deviate from the normal position when wearing the eye control device, which causes the situation to be out of bounds during image capture and conversion, the present invention can be corrected for the deviation of the user's wearing position.

[3] can avoid the error of natural vibration of the eyeball. In the image processing of the eye control system, since the human eye itself has an involuntary tremor phenomenon, the present invention can use the same cumulative number of moving directions to filter out the offset caused by natural vibration, so that the entire output The trajectory is close to the actual value, so it can improve the handling and stability of the subsequent eye control system, [4] can adjust the sensitivity of the system. It can adjust the preset threshold k of the same cumulative number of movement directions to slightly adjust the sensitivity of the system. The present invention has been described in detail with reference to the preferred embodiments thereof, without departing from the spirit and scope of the invention. 19 1261785 From the above detailed description, it will be apparent to those skilled in the art that the present invention can achieve the foregoing objects and is in accordance with the provisions of the Patent Law.

20 1261785 [Simple description of the diagram] The first diagram is the schematic diagram of the wearing of the head-mounted device. The second diagram is the schematic diagram of the angle of the image of the head-mounted device. The third picture is the image-taking diagram of the large-scale projection surface. The schematic diagram of the eye image shows the top, right, middle, left and lower reference points on the target screens of the fifth, fifth, fifth, fifth, and fifth E images.

The points above the gaze (PQ, Q.) and the right points (P, Q,) are displayed on the eye images captured by the sixth, sixth, sixth, sixth, and sixth E systems. Look at the central point (P2, Q2), look at the left point (P3, Q3), and look at the lower point (p4, Q4). The seventh figure is the point in the conventional five-point correction step. Schematic diagram of oblique axis angle and proportional relationship FIG. 9 is a flow chart of the present invention. FIG. 10 is a schematic diagram of a screen of the present invention in a five-point calibration step.

11 is a schematic diagram showing the correspondence relationship between the positions of the eyeballs of the present invention. FIG. 12 is a schematic diagram showing an exemplary embodiment of the present invention. FIG. 14 is a schematic view showing the object of observation of the present invention. FIG. Comparison Figure 2 0 (Description of the figure) Five-point correction step 1 1 Correction coefficient setting step 1 2 Quadrant discrimination step 1 3 Out-of-grid offset determination step 1 4 Coordinate conversion step 1 5 Obtain complex moving point Step 1 6 Determine the moving direction Step 1 7 Output Track Calculation Step 1 8 21 1261785 10 15 20

Right correction point (G χ 1, G y 1 ) Left correction point (G x 3, G y 3 ) Screen coordinates (Gx, Gy) First quadrant midpoint coordinates (S x 1, S y 1 ) in the second quadrant Point coordinates (S x 2, S y 2 ) third quadrant midpoint coordinates (S x 3 , S y 3 ) fourth quadrant midpoint coordinates (S x 4, S y 4 ) fixation points P0, PI, P2, P3 , P4 > P5, P6, P7, P8, P9, P 1 0 Area A, B, C Target screen 9 1 Eye image 9 3 Reference point C0 (X., Y0) Reference point C2 (X3, Y3) ) Lower reference point C4 (X4, Y4) Look at the upper point (Pg, Q〇) Look at the center point (p2, q2) Look at the lower point (p4, q4) Gaze point (Μ X, M y) Horizontal axis value / w Correspondence Coordinate (Dm, Dn) Look at the midpoint (Μ 0 x, Μ 0 y) Look at the upper point (Μ x 2, M y 2 ) Look at the correction point in the lower point (Μ x 4, M y 4 ) (S c χ, S cy ) Upper correction point (Gx2, Gy2) Lower correction point (G x 4, G y 4 ) Eye image capturing device 9 2 Tilt angle θ Right reference point CUX^Y!) Left reference point C 3 ( X 3 , Y 3) Look at the right point (P !, Q !) Look at the left point (P 3, Q 3) The coordinates have been captured (pm, pn) The vertical axis value is the right point for a long time (Μ χ 1, M y 1 Gaze Point (Μ χ 3, M y 3) moving direction vQ1, v2, V3, v4, v5, v6, v7, v8, v9, 22

Claims (1)

1261785 \ Pickup, patent application scope: 1. An eyeball trajectory tracking method, comprising the following steps: 1. Five-point calibration step: The user separately looks at the correction point 5 (S c X, S cy ) on a display screen Right correction point (GX 1, G y 1 ), upper correction point (Gx2, Gy2), left correction point (Gx3, Gy3), and lower correction point (GX 4, G y 4 ), and through an eye image撷Take the device to capture the image of the eye and get the midpoint of the eye (Μ Ox, Μ 0 y ), gaze at the right point (Μ x 1,M y 1 ), gaze at the upper point (Μ x 2 10 , M y 2 ), gaze at the left point (Μ χ 3 , Μ y 3 ), and gaze at the next point (Μ χ 4 , My4); Second, the correction coefficient setting steps: Set four sets of correction coefficients (Α, Α), ( /? γ2, household v2), ( uv3), (β χΑ ^ β } ^); step: When a user's gaze point (Μ χ, M y ) is obtained, it can be mapped to a screen coordinate (Gx, Gy); by the gaze point (Mx, My) and the gaze midpoint (Μ 0 χ, Μ 0 y ) The positive and negative values after subtraction are (positive, positive), (negative, positive), (negative, negative), (positive, negative) and respectively judged to be 2 0, the first quadrant, the second quadrant, the first Three quadrants, fourth quadrant four, out of the offset determination step: If the number of outbounds per second occurs more than a predetermined number of times, it can be determined that the user's wearing position is biased, that is, the fixation point (Mx, My) 23 1261785 Exceeding the correction range, and calculating the following formula: If Mxs>Mx, Bay 1J Ax = Mx-Mxs, No Bay 1J Αχ 2 0; If Mys > M y, Bay 1J A y = M y - M ys, No Bay 1J A y = 0 ; where Mxs and Mys are the boundary points of the quadrant, and the maximum out-of-place position offset Αχ and Ay can be calculated; 5. Coordinate conversion step: If it is the first quadrant, then The screen coordinates (GX, G y ) are 4 2 1+/'-丨+A with \ 丨) Gy 2 σ 々丨 Α + Αι (from less + ♦- 夕丨, 丨) 10 Wherein, the coordinates of the first quadrant are vl t:v G y1 - s, and wherein the coordinates of the first quadrant of the eyeball position are 1λΊ Μ, and Μχ'-Μα 2 is 1 少 less, 2 〇y 15 In the second quadrant, the screen coordinates (G x , G y ) are Gx=Scx^Ix2+Px2{Mx^Ax-Sx2) Gy = Scy + ^.V2 + βy2 (My + ^ ~ Sy2 ) where The coordinates of the midpoint of the second quadrant are x3, Svl = Sr 2 0 and wherein the coordinates of the second quadrant of the eyeball position are 24 1261785 to 10, and if the third quadrant is, the screen coordinates (GX, G y ) Gx = ^cv + 1x3 + β <3 (ΜΧ + Αχ- Sx3 ) Gy=Scy+Iy3^-^y3(My+Ay-Sv3) wherein the third quadrant has a point coordinate of y4 and wherein The third quadrant midpoint coordinates 'M,, Mox-Mx 2 •Μ', Μ, ',, Μ less 4 2 If the fourth quadrant, the screen coordinates (Gx, Gy) are Gx=Scx +Ix^PxA(Mx^Ax-SxA)g less =~+仏+eight 4 + ♦ _~) where the coordinates of the fourth quadrant are 15 and wherein the coordinates of the fourth quadrant of the eyeball position are M, Μχ'-Μυ 2 [., νΐ Μ, ' Μ,, ', a Μ ν4 2 2 · If you can, you can quickly Calculate the eyeball tracking method (G χ, G y ) 眼 The eyeball tracking method described in item 1 of the patent application scope includes the following subsequent processing steps: 25 1261785 6. Steps of obtaining a plurality of moving points: Moving the eyes of the user A predetermined trajectory is sampled at different times (t) to obtain a plurality of moving points (X (t), y (t )), and two adjacent moving points form a motion vector, x (〇y{t) VII. Discriminate the moving direction step: Also, 11^(011=^/^(0)2 + (7(0)2 , when ||v(r-,)|| ^-|||((|||| (r-,)|| ^ 10 Among them, the Δ household is the user-defined error value, as long as it conforms to the above formula, we call the same moving direction, and define η as the cumulative number of movements with the same moving direction; 8. Output trace calculation step: When the cumulative number η is less than Equal to or greater than a preset threshold k , the output track 圮 (0 can be calculated by the following formula: 15 v(t) -v(t - Ϊ) v〇(t) = \m(kn)\\v(t)\\ if n <k if n&gt ;k where m is a predetermined correction factor, whereby the eyeball movement output trajectory 圮(1) can be avoided by avoiding the error of natural vibration of the eyeball.