SGS - Acquiring the Kinematic Data

To begin the process of obtaining the kinematic data, we first needed a recording of athletes running during a competitive sprinting event. Our group consulted Dr. Larry Abraham of the Motor Coordination Lab at UT to ask to borrow footage that had been previously discussed in his course, KIN 395 - Biomechanics of Sport. The recording is of a women's race at a college track meet, presumably the 4 x 400 meter relay. The third runner to enter the frame of the video was selected for analysis, as she had the clearest complete gait cycle. The leg facing the camera (the athlete's left) was chosen for analysis. The relevant clip of the runner moving through a single gait cycle is shown in Figure 2.

Figure 2. Clip from original footage used to obtain kinematic data

In order to measure and track the position of the athlete's hip, knee, and ankle through one gait cycle, we made use of the motion analysis software called Dartfish. This software was provided courtesy of Dr. Abraham and the Kinesiology PhD student Dhruv Gupta, who teaches its use in a course on biomechanics. Dartfish allows the user to record the position of objects in a video by dropping markers and moving them frame-by-frame. While motion capture using MATLAB is possible, it is normally done using color video and bright reflective markers to allow for easier tracking. The frame-by-frame method with the Dartfish interface was the more reliable option given the quality of our video, and successfully output the x-y coordinate trajectories of the hip, ankle, and knee. The raw data results are shown in Figure 3a. The axes are not shown strictly to scale in order to exaggerate the vertical motion of the joints.

To find the path of the ankle relative to the pelvis, we subtracted the hip coordinates from the ankle, producing the gait cycle shown in Figure 3b. The exact measurements of this data are unknown, as there were no reliable reference objects of known length in the original video. However, this is ultimately unimportant as we had already planned to make a scale model that reproduces this motion and not an exact replica. For this reason, no units are included in either Figure 3a or 3b.

An important detail of this process is that a single measured gait cycle can be unreliable if the goal is to accurately reproduce the athlete's movement. The preferred method is to record multiple gait cycles and average them together to account for slight deviations from cycle to cycle and any noise in the collected data. However, as this is more of a "proof of concept" type of project, we decided to attempt to reproduce the sprinting gait cycle. The results from modeling and optimizing our robot demonstrated that even fitting to noisy data such as this was possible. As will be seen however, they are not perfect, which can partly be attributed to relying on a single set of data.

Figure 3a. Ankle trajectory as measured by Dartfish.

Figure 3b. Ankle trajectory relative to hip of sprinter during one full gait cycle.

The x- and y-axes represent relative position and are unitless.