Eye tracker system and methods for detecting eye parameters
An improved eye tracker system and methods for detecting eye parameters including eye movement using a pupil center, pupil diameter (i.e., dilation), blink duration, and blink frequency, which may be used to determine a variety of physiological and psychological conditions. The eye tracker system and methods operates at a ten-fold reduction in power usage as compared to current system and methods. Furthermore, eye tracker system and methods allows for a more optimal use in variable light situations such as in the outdoors and does not require active calibration by the user.
1. A system for tracking parameters of an eye, the system comprising:
an imaging component facing at least said eye and positioned to image a pupil of said eye;
a controller operably connected to the imaging component, the controller performing the steps of:
obtaining a subsampling of a set of pixels of a pupil image gathered from said imaging component;
estimating, based on said subsampled set of pixels, a location parameter based on a center of said pupil according to a x coordinate and a y coordinate and a size parameter according to a radius r of said pupil, wherein said estimating step further comprises the step of using an artificial neural network prediction model; and
tracking said pupil by sampling a row and a column of pixels based on said estimated location parameter and said estimated size parameter of said pupil, wherein said tracking step further comprises the steps of validating the sampled pixels with the pupil image, and using validated sampled pixels to track parameters of the eye.
2. The system of claim 1 wherein said tracking step performed by the controller further comprises the steps of: median filtering said row and said column of pixels, detecting regions of said eye using edge detection, finding a midpoint of said row and said column of pixels, performing a validity check to determine said midpoint is consistent, and if the validity check shows an error, said controller repeats said estimating step and said tracking step.
3. The system of claim 1 further comprising a computing device, wherein the controller further comprises the steps of: calibrating the artificial neural network by capturing images of said eye in at least one lighting condition; communicating by the controller said images to the computing device, wherein said computing device performs said estimating step and said tracking step to form a signature of said eye in said at least one lighting condition.
4. The system of claim 1 further comprising an infrared photodiode that detects at least one lighting condition.
5. The system of claim 1 further comprising an infrared illuminator positioned to illuminate said eye.
6. The system of claim 3 wherein said calibrating step is completed automatically.
7. The system of claim 1 wherein said controller omits said tracking step when at least one lighting condition is outdoors and said controller performs the step of increasing a quantity of said subsampled set of pixels of said obtaining step.
8. A method for tracking parameters of an eye comprising the steps of:
providing a camera facing at least said eye and positioned to capture an image of a pupil of said eye;
obtaining by a controller a subsampling of a set of pixels gathered from the image;
estimating by the controller, based on said subsampled set of pixels, a location parameter based on a center of said pupil according to a x coordinate and a y coordinate and a size parameter according to a radius r of said pupil, wherein said estimating step further comprises the step of using an artificial neural network prediction model;
tracking by the controller said pupil by sampling a row and a column of pixels from the image based on said estimated location parameter and said estimated size parameter, wherein said tracking step further comprises the steps of validating the sampled pixels with the image, and using validated sampled pixels to track parameters of the eye.
9. The method of claim 8 wherein said tracking step further comprises median filtering said row and said column of pixels, detecting regions of said eye using edge detection, finding the midpoint of said row and said column of pixels, performing a validity check to determine said midpoint is consistent, and if the validity check shows an error, said system repeats said estimating step and said tracking step.
10. The method of claim 8 further comprising the step of calibrating by the controller the artificial neural network prediction model by capturing images of said eye in at least one lighting condition; communicating said images to a computing device, wherein said computing device performs said estimating step and said tracking step to form a signature of said eye in said at least one lighting condition.
11. The method of claim 10 further comprising the step of hard-coding by the controller said signature into said controller for use in said estimating step.
12. The method of claim 11 further comprising the step of detecting by an infrared photodiode said at least one lighting condition.
13. The method of claim 8 further comprising the step of illuminating by an infrared illuminator said eye.
14. The method of claim 10 wherein said calibrating step is completed automatically.
15. The method of claim 8 wherein said tracking step is omitted when said at least one lighting condition is outdoors and further comprising the step of increasing a quantity of said subsampled set of pixels of said obtaining step.
16. The system of claim 1 wherein the artificial neural network prediction model is:
X
^
=
B
x
O
+
∑
k
=
1
K
W
kx
WO
H
k
Y
^
=
B
y
O
+
∑
k
=
1
K
W
ky
HO
H
k
R
^
=
B
r
O
+
∑
k
=
1
K
W
kr
HO
H
k
H
k
=
tanh
(
B
k
H
+
∑
i
=
1
D
∑
j
=
1
D
W
ijk
IH
I
ij
)
I ij represents a pixel at row i and column j of the pupil image,
H k represents K hidden units,
W ijk IH represents input-to-hidden parameters for each pixel location (i,j) and each hidden unit H k ,
B K H represents a hidden unit bias parameter for each hidden unit H k ,
W kx HO and W ky HO represent hidden-to-output parameters mapped between hidden unit H k and the x coordinate and the y coordinate and the radius r, and
B x O , B y O , B r O represent output bias parameters for the x coordinate and the y coordinate and the radius r.
17. The method of claim 8 wherein the artificial neural network prediction model of said using step is:
X
^
=
B
x
O
+
∑
k
=
1
K
W
kx
WO
H
k
Y
^
=
B
y
O
+
∑
k
=
1
K
W
ky
HO
H
k
R
^
=
B
r
O
+
∑
k
=
1
K
W
kr
HO
H
k
H
k
=
tanh
(
B
k
H
+
∑
i
=
1
D
∑
j
=
1
D
W
ijk
IH
I
ij
)
I ij represents a pixel at row i and column j of the pupil image,
H k represents K hidden units,
W ijk IH represents input-to-hidden parameters for each pixel location (i,j) and each hidden unit H k ,
B K H represents a hidden unit bias parameter for each hidden unit H k ,
W kx HO and W ky HO represent hidden-to-output parameters mapped between hidden unit H k and the x coordinate and the y coordinate and the radius r, and
B x O , B y O , B r O represent output bias parameters for the x coordinate and the y coordinate and the radius r.