‹date/time›=
‹footer›
‹#›
------=_NextPart_01C60EFA.1ADBE7B0
Content-Location: file:///C:/272B5D12/GaitEventDetectionPoster_files/pres.xml
Content-Transfer-Encoding: quoted-printable
Content-Type: text/xml; charset="utf-8"
![](3D"slide0001_background.gif")
![](3D"slide0001_image001.gif")
Real-Time Gait E=
vent
Detection from
Contributors: AC Davison, S Dong, S Farquhar
Advisors: D Reisman, K Rudolph, JQ Sun
Knee Brace Integrated Kinematic Sensors<=
/span>
![](3D"slide0001_image002.gif")
Motivation
A Smart Knee Brace (SKB) is being developed for <=
/span>stroke patient gait rehabilitation. The specific =
function of the SKB is to regulate hemiplegic walking by providing support in the stance phase,=
while allowing free knee motion in the swing phase. The
SKB has led to the need for a real-time
gait phase detection algorithm based solely on sensors directly attached to the brace.=
=
Methods
•<=
/span>Kinematic data was collected using motion =
capture techniques. An accelerometer was <=
/span>attached to the shank to collect the horizontal <=
/span>component of the acceleration.
=
•
•<=
/span>In real time, this data will be measured using gyroscopes and a knee angle sensor.
•
•<=
/span>The algorithm utilizes the negative e=
xtreme
of the
accelerometer to detect the initial contact (IC) of the foot (Figure 1). To detect the toe-off (TO), the minimum of the knee velocit=
y is
used in
conjunction with an increasing knee angle (Figure 2).
•
•<=
/span>The algorithm requires an initial
calibration of three
complete gait cycles. During=
the calibration, the algorithm detects the
critical valu=
es
after they occur using global
extrema. To account for variations in walking =
speed,
the algo=
rithm
continually updates the history of the critical values.
![](3D"slide0001_image004.gif")
![](3D"slide0001_image006.gif")
![](3D"slide0001_image007.gif")
References
MH Granat and K Tong (1998) =
Medical Engineering and
Physics. 21, 87-94. "A practical Gait
Analysis System Using Gyroscopes“
HJ Chizeck and MM Skelly (20=
01) IEEE Transactions on Neur=
al
Systems and Rehabilitation Engineering. 9(1), 59-67. "Real-Time Gait Event Detection for Paraplegic FES
Walking"
=
Results
•<=
/span>The algorithm was applied to
healthy patients and a stoke patient.&nb=
sp;
The algorithm successfully identified the different gait cycles with high accuracy in compar=
ison
to foot s=
witch
data.
•
•<=
/span>Typical results for a stroke
patient are shown in Figure 3.
Table 1 summarizes the error with different patients.
Table 1: Error (in samples) of
Prediction at 120Hz
![](3D"slide0001_image009.gif")
![](3D"slide0001_image010.gif")
=
Further Study
=
•<=
/span>The algorithm needs further
validation using stroke patients with different gait styles.
<=
/span>
•
•<=
/span>Gyroscopes need to be
implemented in real-time.
•
•<=
/span>Eventually a microprocessor =
will
be used to run the algorithm in real-time.
0.788
0.55
1.12
1.53
0.62
3
% of Cycle
1.45
1.1
1.78
2.17
2.23
2.14
STD
1.34
0.94
1.9
-2.6
-1.05
-5.11
Mean
Initial Contact
Toe Off
Initial Contact
Toe Off
Initial Contact
Toe Off
Normal Subject
Free Speed
Stroke Patient
Constant Speed
Stroke Patient
Free Speed