Conclusiones del capítulo 7

For each subject and WF/WB trial, the performed gait cycles were automatically detected using Kinect data by relying on our gait cycle detection algorithm (same window sizes as those used in the study presented in the previous chapter). The corresponding actual gait cycles were automati- cally identified using Qualisys data relying on the method described in Appendix C.2.

For each gait cycle, we computed the 13 spatiotemporal gait parameters (i.e., parameters involving time and/or distance) included in Table 6.2, and the 8/9 (Kv1/Kv2) kinematic gait param- eters (i.e., parameters involving angles) included in Table 6.3. For each parameter, we then obtained the mean value per subject, when taking into account N gait cycles (chosen randomly if the number of gait cycles for a subject is higher than N).

For Qualisys, the 3-D position of each body joint used for gait parameter computation was calculated from the 3-D position of the relevant markers, as indicated in Table 6.4. The 3-D joint data were then processed using a zero-lag low-pass fourth order Butterworth filter with a cut-off fre- quency of 15 Hz. The filter parameter values were chosen by taking into account the frequency con- tent of the signals.

For Kinect, the velocity, distance and angle signals were processed using a zero-lag low-pass Butterworth filter. For each gait parameter, we explored the filter order values of 2, 4 and 6, and cut- off frequency values between 1 and 9 Hz inclusive (integer values). We then chose the value for the filter parameters that minimized the mean absolute error between Kinect and Qualisys.

For each gait parameter, we excluded the gait cycles with outliers for the associated measure obtained from Qualisys data. For velocity and distance measures, we considered as outliers the frames with value above 5 m/s and 2 m, respectively. For all measure types, we also considered as outliers the frames for which the value is more than 40 scaled median absolute deviations away from the median for the following signal: sum of absolute difference of the joint position between consec- utive frames in each axis.

All signal processing and analysis, including outlier detection, filter optimization, gait cycle detection and gait parameter computation, were performed in Matlab (version R2015a) [234].

6.2 Materials and Methods

Table 6.2. Spatiotemporal gait parameters computed over 3-D joint data, including a description and the equation used for its computation.

Spatiotemporal

gait parameter Definition Equation

Stride duration Duration of the gait cycle or stride. (5.11) Step duration Duration of the left/right step, when considering a

left/right gait cycle. (5.12)

Stance

duration Duration of the stance phase. (5.13)

Swing duration Duration of the swing phase. (5.14)

Single support

duration Duration of the single-limb support phase. (5.15) Double support

duration Duration of the double-limb support phase. (5.16)

Stride length

Distance (in the xz-plane) between the position of the left/right ankle at the beginning and end of the

left/right gait cycle. The distance is computed using (3.6), where P1 and P2 correspond

to the position of the ankle (in the xz-plane) at the two considered

instants. Step length

Distance (in the xz-plane) between the position of the left/right ankle at the instant during the swing phase when the ankle distance is minimum, and at the instant corresponding to the end of the left/right

gait cycle.

Step width Minimum of the ankle distance (in the xz-plane) _{during the swing phase.}

The ankle distance is computed for each frame using (3.6), where

P1 and P2 correspond to the

position of the left and right ankle (in the xz-plane), respectively.

Gait speed

Mean of the trunk velocity (mean of velocity of the head, neck, spine shoulder, spine middle, and spine

base joints).

The velocity of each joint is computed for each frame using

(3.4). Gait speed

variability

Standard deviation of the trunk velocity (mean of velocity of the head, neck, spine shoulder, spine

middle, and spine base joints).

Foot swing velocity

Maximum of the left/right ankle joint velocity during the swing phase, when considering a

left/right gait cycle.

Arm swing velocity

Maximum of the left/right hand velocity (mean of velocity of the left/right wrist and hand joints)

during the stance phase, when considering a left/right gait cycle.

Table 6.3. Kinematic gait parameters computed over 3-D joint data, including a description and the equation used for its computation.

Kinematic gait

parameter Definition Equation

Neck angle Mean of the angle at neck (defined by the head, neck and spine base _joints).

The angle was computed for

each frame using (3.7). Spine shoulder

anglea

Mean of the angle at spine shoulder (defined by the head, spine shoulder and spine base joints).

Spine middle angle

Mean of the angle at spine middle (defined by the head, spine middle and spine base joints).

Elbow angle minimum

Minimum of angle at the left/right elbow during the stance phase, when considering a left/right gait cycle. The angle at the elbow is defined by

the corresponding shoulder, elbow, and wrist joints. Elbow angle

maximum

Maximum of angle at the left/right elbow during the stance phase, when considering a left/right gait cycle.

Knee angle minimum

Minimum of angle at the left/right knee during the swing phase, when considering a left/right gait cycle. The knee angle is defined by the

corresponding hip, knee, and ankle joints. Knee angle

maximum

Maximum of angle at the left/right knee during the stance phase, when considering a left/right gait cycle.

Hip angle range

Difference between maximum and minimum value of the left/right hip angle, during the stance and swing phases respectively, when considering a left/right gait cycle. The hip angle is defined by the corresponding knee joint, hip joint, and point with knee y-coordinate

and hip x- and z-coordinates.

Ankle angle range

Difference between maximum and minimum value of the angle at the left/right ankle, during the swing and stance phases respectively, when

considering a left/right gait cycle. The angle at the ankle is defined by the corresponding knee, ankle, and foot joints.

a _{Kinect v2 only}

Table 6.4. Computation of the 3-D position of joints from Qualisys data. Joint(s) tracked by Kinect Joint(s) position computed from Qualisys markersa

Head P = PHEAD

Neck P = PNECK

Spine shoulder P = PIJ

Spine middle P = PPX

Spine base P = PHIP CENTER

Left and right shoulder Pleft = PLAC andPright = PRAC

Left and right elbow Pleft = (PLLELB + PLMELB)/2bandPright = (PRLELB + PRMELB)/2b

Left and right wrist Pleft = (PLULN + PLRAD)/2bandPright = (PRULN + PRRAD)/2b

Left and right hand Pleft = PLHAND andPright = PRHAND

Left and right hip Pleft = PLASIS andPright = PRASIS

Left and right knee Pleft = (PLLK + PLMK)/2bandPright = (PRLK + PRMK)/2b

Left and right ankle Pleft = (PLLA + PLMA)/2bandPright = (PRLA + PRMA)/2b

Left and right foot Pleft = PLFM2 andPright = PRFM2

6.2 Materials and Methods