Device and method for detecting drowsiness using eyelid movement
View Patent ↗A drowsiness detection device and method for a person ( 4 ), having a digital camera device ( 2 ) and a control unit ( 5 ). The digital camera device ( 2 ) includes a video processor ( 6 ). The digital camera device ( 2 ) is arranged to detect the eyes ( 3 ) of the person ( 4 ) in a series of images. The video processor ( 6 ) is arranged to detect eyelid movements in the images and convert images of the eyelid movements to an electric signal (s(t)) which is transferred to the control unit ( 5 ) The control unit ( 5 ) is arranged to fit a modeled signal (y(t)) against the electric signal (s(t)) and to determine whether the modeled signal (y(t)) constitutes an eyelid blink. The control unit ( 5 ) uses a number of modeled signals (y(t)) determined to constitute an eyelid blink to determine whether the level of alertness has fallen below a certain threshold.
1. A method for detecting the level of alertness of a person, the method comprising the steps:
detecting eyelid movements of the person by a camera in communication with a control unit;
creating an electric signal (s(t)) from the camera that represents the detected eyelid movements;
analyzing the electric signal (s(t)) by the control unit by fitting a modeled signal (y(t)) against the electric signal (s(t)), the modeled signal (y(t)) is defined as:
y
(
t
)
=
{
k
1
·
log
(
t
2
-
t
)
+
m
1
for
t
1
≤
t
<
t
2
m
2
for
t
2
≤
t
<
t
3
k
3
·
log
(
t
-
t
3
)
+
m
3
for
t
3
<
t
≤
t
4
where k 1 , k 3 , m 1 , m 2 and m 3 are constants that are defined for each eyelid blink and the times t 1 , t 2 , t 3 , t 4 define time intervals of the modeled signal (y(t));
determining by the control unit whether the modeled signal (y(t)) constitutes an eyelid blink; and
determining by the control unit, based on the analyzing step, whether the level of alertness of the person falls below a certain threshold using a plurality of the modeled signals (y(t)) when each of the modeled signals has been determined to constitute the eyelid blink.
2. A method according to claim 1 , further comprising that the control unit fitting a modeled signal (y(t)) against the electric signal (s(t)) comprises the steps of:
calculating an error ε according to
ɛ
=
1
N
·
∑
i
=
sample
(
t
1
)
sample
(
t
4
)
s
(
i
)
-
y
(
i
)
such that the times t 1 , t 2 , t 3 , t 4 defining the time intervals and the constant m 2 are varied to minimize the error ε; and estimating the other constants k 1 , k 3 , m 1 and m 3 using regression.
3. A method according to claim 1 further comprising that the electric signal (s(t)) is filtered by the control unit in order to reduce noise.
4. The method for detecting the level of alertness of the person in accordance with claim 1 further comprising the step of detecting eyelid movements of the person includes capturing a plurality of images of the eyes.
5. The drowsiness detection device for a person in accordance with claim 4 further comprising the digital camera device captures a plurality of the images of the eyes.
6. A drowsiness detection device for a person, the drowsiness detection device comprising a digital camera device and a control unit is configured to fit a modeled signal (y(t)) against an electric signal (s(t)) from the camera device and to determine whether the modeled signal (y(t)) constitutes an eyelid blink, where the control unit further is arranged to use a plurality of the modeled signals (y(t)) that each has been determined to constitute the eyelid blink in order to determine whether the level of alertness of the person has fallen below a certain threshold, the modeled signal (y(t)) is defined as:
y
(
t
)
=
{
k
1
·
log
(
t
2
-
t
)
+
m
1
for
t
1
≤
t
<
t
2
m
2
for
t
2
≤
t
<
t
3
k
3
·
log
(
t
-
t
3
)
+
m
3
for
t
3
<
t
≤
t
4
where k 1 , k 3 , m 1 , m 2 and m 3 are constants that are defined for each eyelid blink and the times t 1 , t 2 , t 3 , t 4 define time intervals of the modeled signal (y(t)), the digital camera device including a video processor, where the digital camera device is arranged to detect at least one eye of the person in a series of images, and the video processor is arranged to detect eyelid movements in the images and convert images of the eyelid movements to an electric signal (s(t)) which is transferred to the control unit).
7. A drowsiness detection device according to claim 6 , further comprising that the control unit is arranged to filter the received electric signal such that the control unit is arranged to fit the modeled signal (y(t)) against the filtered electric signal (s(t)).