Estimation of Center of Mass Trajectory using Wearable Sensors during Golf Swing

Abstract

This study suggests a wearable sensor technology to estimate center of mass (CoM) trajectory during a golf swing. Groups of 3, 4, and 18 participants were recruited, respectively, for the purpose of three validation studies. Study 1 examined the accuracy of the system to estimate a 3D body segment angle compared to a camera-based motion analyzer (Vicon®). Study 2 assessed the accuracy of three simplified CoM trajectory models. Finally, Study 3 assessed the accuracy of the proposed CoM model during multiple golf swings. A relatively high agreement was observed between wearable sensors and the reference (Vicon®) for angle measurement (r > 0.99, random error <1.2° (1.5%) for anterior-posterior; <0.9° (2%) for medial-lateral; and <3.6° (2.5%) for internal-external direction). The two-link model yielded a better agreement with the reference system compared to one-link model (r > 0.93 v. r = 0.52, respectively). On the same note, the proposed two-link model estimated CoM trajectory during golf swing with relatively good accuracy (r > 0.9, A-P random error <1cm (7.7%) and <2cm (10.4%) for M-L). The proposed system appears to accurately quantify the kinematics of CoM trajectory as a surrogate of dynamic postural control during an athlete's movement and its portability, makes it feasible to fit the competitive environment without restricting surface type. Key pointsThis study demonstrates that wearable technology based on inertial sensors are accurate to estimate center of mass trajectory in complex athletic task (e.g., golf swing)This study suggests that two-link model of human body provides optimum tradeoff between accuracy and minimum number of sensor module for estimation of center of mass trajectory in particular during fast movements.Wearable technologies based on inertial sensors are viable option for assessing dynamic postural control in complex task outside of gait laboratory and constraints of cameras, surface, and base of support.

Keywords: Wearable technology; balance; center of mass; dynamic postural control; golf swing; simplified biomechanical model of human body.

Figures

Figure 1.

Experimental setup for assessing the accuracy of the sensor in estimating 3D segment angle. Four markers were positioned on a fixed platform and attached to the participant’s lower back to estimate lower back motion in 3D dimension and compare the results with those obtained from the sensor.

Figure 2.

Estimation of CoM using camera based system (Vicon). (a) A set of 36 reflective markers were positioned on the participant to estimate CoM as suggested by Vicon’s "Plug-in Gait full body modeling" with a single sacral marker. (b) Trajectory of CoM was estimated using "Plug-in Gait full body modeling".

Figure 3.

Sensor attachment for estimation of CoM. Three sensors were attached on the shank (a), thigh (b), and lower back (c), respectively. This configuration allows comparing the accuracy of one-link, two-link, and three-link model of human body for estimating CoM trajectory during athletic movements while standing (e.g., golf swing).

Figure 4.

Simplified biomechanical human body model for estimating CoM trajectory. (a) one-link model: in this model we assumed that CoM is equivalent to lower back motion. (b) two-link model: In this model we assumed human body as two rigid segments, one rotating around ankle and the other one rotating around hip joint. (c) three-link model: in this model we modelled human body as three rigid segments rotating around ankle, knee, and hip joints.

Figure 5.

Estimated lower back angle by Vicon (dash line) and inertial sensors (solid line) for a typical participant and during movements in sagittal plane.

Figure 6.

Comparison between estimated values using 2- and 3-segment models as well as COP values measured using pressure platform. CoM values estimated by full body model using Vicon for (A) anterior-posterior and (B) medial-lateral direction.

Figure 7.

(A) Maximum distance of CoM estimation during swing per subject using sensor-derived values versus Vicon derived values. (B) Bland-Altman plot for two methods of CoM estimation.