Calibration and Poin Software Design

TELKOMNIKA Vol. 9, No. 2, 382

2.3.2. Calibration and Poin

Calibration is necess rectangle while the human ey center of the pupil to the coo system is restarted due to var the screen every time the sam shapes will require a new cal 16, or 25 pixel locations, by a can be shown in Figure 7. a Figure 7. Calibrat By using the center lo coefficients of the calibration [8], [20], [21]. This paper used equation 5 and 6 order technique used to approximat formula: 1 ∗ 1 ∗ 1 ∗ 4 ∗ 1 ∗ 4 ∗ 1 ∗ 4 ∗ 7 ∗ ∗ 1 ∗ 4 ∗ 7 ∗ ∗ , and , represen respectively. With 9,16, or 25 25 equations for each of thi coefficients is used least squa 7 and 8 here only shown s , August 2011 : 377 – 386 int Transformation ssary due to the fact that a screen monitor is a eye is not. Mapping is required to transform the c oordinates on the display. Calibration must be don variations in use. The eye will not be in the same lo ame user wears it, also, different users with differ alibration. In this research, the calibration process y asking the user to ‘look at the dot’ in monitor. Th b c ration pixel locations, a 9 pixel, b 16 pixel, c 25 location of the eye when looking at those known p n equation are determined using polynomial regre ed first equation 1 and 2, second equation 3 der polynomial regression. Polynomial regression ate correlation of variables. Shown below the polyn 2 ∗ 3 2 ∗ 3 2 ∗ 3 ଶ ∗ ∗ 5 ଶ ∗ 6 2 ∗ 3 ଶ ∗ ∗ ∗ 5 ଶ ∗ 6 2 ∗ 3 ଶ ∗ ∗ 5 ଶ ∗ 6 ଷ ∗ 8 ∗ ଶ ∗ 9 ଷ ∗ 10 2 ∗ 3 ଶ ∗ ∗ 5 ଶ ∗ 6 ଷ ∗ 8 ∗ ଶ ∗ 9 ଷ ∗ 10 ents center pupil coordinate and target screen mo 25 points sampletraining points in the monitor will this 1, 2, 3, 4, 5 and 6 equations. To o uare method. This method will convert the equatio n sample for first order polynomial from equation 1 a ISSN: 1693-6930 a flat n by m pixel coordinates of the one every time the location relative to ferent eye and face ss is done using 9, The pixel locations c 25 pixel n pixel locations the ression similar with 3 and 4, or third ion is a statistical lynomial regression 1 2 3 4 5 6 monitor coordinate ill produce 9,16, or o obtain regression tions in matrix form 1 and 2. TELKOMNIKA ISSN: 1693-6930 Low-Cost Based Eye Tracking and Eye Gaze Estimation I Ketut Gede Darma Putra 383 n ∑ Px ୧ ୬ ୧ୀଵ ∑ Py ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ ∑ Px ଶ ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Py ୧ ∑ Py ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Py ୧ ∑ Py ଶ ୧ ୬ ୧ୀଵ KoefX1 KoefX2 KoefX3 = ∑ Sx ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Sx ୧ ∑ Py ୧ ୬ ୧ୀଵ Sx ୧ 7 n ∑ Px ୧ ୬ ୧ୀଵ ∑ Py ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ ∑ Px ଶ ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Py ୧ ∑ Py ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Py ୧ ∑ Py ଶ ୧ ୬ ୧ୀଵ KoefY1 KoefY2 KoefY3 = ∑ Sy ୧ ୬ ୧ୀଵ ∑ Px ୧ ୬ ୧ୀଵ Sy ୧ ∑ Py ୧ ୬ ୧ୀଵ Sy ୧ 8 KoefY1, KoefY2, KoefY3 are the regression coefficients that will be computed. The regression coefficients are calculated using Gauss Elimination methods. The next step after the coefficients are obtained is transform or mapping the pupil coordinate to screen coordinate to get the gaze point Gx,Gy . This point can be achieved by multiplication of pupil coordinate output from detection of center pupil with regression coefficients output from calibration process. The first equation 9 and 10, second equation 11 and 12, and third equation 13 and 14 order polynomial gaze point can be computed as below: Gx = KoefX[0] + Px ∗ KoefX[1] + Py ∗ KoefX[2] 9 Gy = KoefY[0] + Px ∗ KoefY[1] + Py ∗ KoefY[2] 10 = [0] + ∗ [1] + ∗ [2] + ଶ ∗ [3] + 11 Gy = KoefY[0] + Px ∗ KoefY[1] + Py ∗ KoefY[2] + Px ଶ ∗ KoefY[3] + Px ∗ Py ∗ KoefY[4] + Py ଶ ∗ KoefY[5] 12 Gx = KoefX[0] + Px ∗ KoefX[1] + Py ∗ KoefX[2] + Px ଶ ∗ KoefX[3] + Px ∗ Py ∗ KoefX[4] + Py ଶ ∗ KoefX[5] + Px ଷ ∗ KoefX[6] + Px ଶ ∗ Py ∗ KoefX[7] + Px ∗ Py ଶ ∗ KoefX[8] + Py ଷ ∗ KoefX[9] 13 Gy = KoefY[0] + Px ∗ KoefY[1] + Py ∗ KoefY[2] + Px ଶ ∗ KoefY[3] + Px ∗ Py ∗ KoefY[4] + Py ଶ ∗ KoefY[5] + Px ଷ ∗ KoefY[6] + Px ଶ ∗ Py ∗ KoefY[7] + Px ∗ Py ଶ ∗ KoefY[8] + Py ଷ ∗ KoefY[9] 14 This gaze point can be used to control the movement of mouse cursor using our eye.

3. Results and Analysis