White Paper: Validation Vector T7

Matthew C. Varley, Susanne Ellens, David Carey. Sport, Performance, and Nutrition Research Group, School of Allied Health, Human Services, & Sport, La Trobe University, Melbourne, VIC, Australia.

1. Introduction
2. Methods
3. Statistical Analysis
4. Results
5. Summary
6. References

1. INTRODUCTION

Athlete tracking systems have become an essential tool for sport. These systems allow practitioners to quantify and analyse athletes’ movement to better understand training load, physical performance, tactical behaviour, and injury risk.¹

While many tracking systems are available, most have limitations when used indoors. The ClearSky local positioning system (LPS) is a technology that provides tracking and measurement of athlete performance in indoor environments. Specifically, the system outputs position and position-derived metrics such as distance, velocity, and acceleration.

The Vector T7 is a new wearable device that is used with the ClearSky system. The Vector T7 is similar to previous devices used with ClearSky (Vector S7 and Catapult T6), with a sampling frequency of 10 Hz.

However, the Vector T7 uses a Time Difference of Arrival (TDOA) protocol to derive position whereas the previous devices use a Two Way Ranging (TWR) protocol. The benefit of TDOA protocol is that it requires a substantially lower power consumption for the device compared to TWR protocol while maintaining the accuracy of the positional data.

This reduction in power enables the size of the device to be reduced. The smaller device size allows the Vector T7 to be worn at several positions on the athlete, including traditional placement between the shoulder blades or placed at the waist.

Athlete tracking systems require validation of their ability to measure athlete movement for practitioners to have confidence in the data allowing them to make decisions on training and match practices. As with all technology, manufacturers will release updated models over time, as improvements to both device hardware and their underlying algorithms are made.

Each new model requires validation to determine the ability of the new device to measure what is intended to measure (e.g., position, velocity, and acceleration).¹ This is typically done by comparing the data from the device to a criterion measure. The Vicon system is a motion capture camera system considered to be the gold standard for measuring position. It is common for Vicon to be used as a criterion measure in the validation of athlete tracking technology.^{2, 3}

Studies have assessed the validity of Catapult T6 devices for measuring distance, velocity, and acceleration during team-sport specific tasks, including linear maximal efforts and change-of-direction movements.^4,5,6

These studies used a motion capture camera system (either Vicon or Qualisys Oqus) as a criterion measure with all studies concluding that the Catapult T6 devices had acceptable validity for assessing athlete movement. Given the Vector T7 has only recently been developed, validation of this device is required.

Therefore, the aim of this study was to assess the validity of the new Vector T7 devices for measuring distance, velocity, and acceleration.

→ Download this Vector T7 White Paper by clicking here.

2. METHODS

Six recreationally active male participants (28.8 ± 5.6 years old) participated in this study. All participants provided written consent for their participation in the study and the procedures used were conducted with approval from the Human Research Ethics Committee of La Trobe University.

Data collection was conducted in a sports hall measuring 40 x 70 m, which consisted of three basketball courts. Participants performed seven different movement trials in a 20 x 5 m area on one of the basketball courts. During the trials player movement data was collected via an LPS (Catapult ClearSky, Catapult, Melbourne, Australia) and a 20-camera motion analysis system (Vantage, Vicon Motion Systems, Oxford, UK).

A description and schematic of each trial is shown in Table 1 and Figure 1. Participants were asked to perform all trials at a maximal intensity and to come to a complete stop at the end of the trial (deceleration). Participants performed a 5-min warm-up before the commencement of the trial. Each movement trial was performed twice for a total of 14 trials with each trial separated by ~3 min. 

Participants were equipped with four Vector T7 devices positioned in four different locations. One device was positioned between the participant’s shoulder blades using the manufacturer supplied vest.

The remaining three devices were positioned around the participants waist, specifically at the front (cross section of the midpoint between the two anterior superior iliac spines), left side (cross section of the midpoint between the anterior and posterior superior iliac spines) and back (cross section of the midpoint between the two posterior superior iliac spines) in a custom waistband clip supplied by the manufacturer which clipped on to the participants shorts.

Separate to the movement trials, a static trial was performed to assess the Vector T7 devices for the stability of their positioning using a fixed placement protocol. Three devices were each placed on a tripod (1.5 m elevation) and left to collect data for a 10 min period. Of the devices two were placed on the centre of the middle court and one was placed on the wide edge of the court.

The ClearSky LPS was installed around the sports hall and consisted of 21 anchor nodes fixed at an average height of 8.4 m from the ground with an average distance of 10.4 m between each node. Data was captured at 10 Hz and processed using the manufacturer software (OpenField version 3.9.0). Velocity, acceleration, x-y position and odometer (cumulative distance) data were exported for each trial for further analysis.

The 20-camera motion analysis system (Vicon) sampling at 100 Hz, was used as the criterion distance, velocity, and acceleration measure. The cameras were mounted on tripods and placed 3 m from the perimeter of the area where the movement trials were performed. Four retro-reflective markers with a diameter of 32 mm were placed on the outside of the manufacturer supplied vest and each waistband clip containing the Vector T7 devices, in correspondence with the middle of each device.  

Vicon data was labelled and processed with Vicon Nexus 2.14. Data processing of raw Vicon data consisted of filtering using a fourth order low pass Butterworth filter with a 3Hz cut-off frequency which was determined based on residual analysis. Gaps in the data ≤50 ms (5 samples) were filled using spline interpolation, gaps ≥50 ms were excluded from analysis. XY-coordinates of the filtered 100 Hz Vicon data were used for analysis, the z-coordinates (vertical displacement) were neglected in the calculations as ClearSky was configured for two-dimensional (2D) positioning.

For each of the four Vicon markers in each movement trial (n=320) the 2D velocity was calculated by differencing the positional data and applying the same filter used in the manufacturer software on the LPS data. This information was provided to the researchers by the manufacturer; however, details are not included here due to the manufacturer’s intellectual property. Similarly, acceleration was calculated by differencing the velocity data and filtering using the manufacturers specifications.

The Vicon derived metrics were down sampled to 10Hz and then synchronised to the Catapult data by cross correlating the velocity signals to find the time offset that maximised the correlation. All data processing and analysis were performed using the R statistical programming language (version 4.0.4)⁷ and the gsignal package⁸.

Table 1. Description of movement trials

Trial #	Description
1	5 m linear sprint
2	10 m linear sprint
3	20 m linear sprint
4	5 m linear sprint into 45-degree Change of Direction into 5m linear sprint
5	5 m linear sprint into 90-degree Change of Direction into 5m linear sprint
6	5 m linear sprint into 180-degree Change of Direction into 5m linear sprint
7	A circuit involving a combination of linear sprints and Change of Directions

Figure 1. A) Schematic of the seven different movement trials. B) Setup during data collection, all trials commenced from the start reference point (white circle), Vicon cameras (black trapezoid), ClearSky anchor nodes (white triangles). Note: Entire Sports Hall (40 x 70m is not shown in full for clarity).

→ Download this Vector T7 White Paper by clicking here.

3. STATISTICAL ANALYSIS

Movement trials

The following metrics were calculated for each trial comparing the Vector T7 and Vicon derived data; root mean standard difference (RMSD) for velocity and acceleration and mean absolute difference (MAD) for the sample-to-sample positional distance. Vicon trials that had greater than 10% missing data were excluded from analyses (n = 12) due to the introduction of edge effects by the filtering approach where gaps existed in the data. Results are presented as mean, median, and interquartile ranges (IQR) for each metric, across all trials, and stratified by device location and movement type.

Static trials

For static device trials the sample-to-sample positional distance and displacement were calculated. Results are presented as the mean, median, IQR, and cumulative total for distance, and the displacement from the first to last sample in the 10-minute trial. The manufacturer metric “odometer” was included in results for each device as a point of reference for calculated distance travelled.

4. RESULTS

Movement trials

Differences between ClearSky LPS and Vicon derived velocity are shown in Table 2 and Figure 2. Of the four device positions tested, the device worn in the vest had the smallest mean and median RMSD, and the device worn at the front of the waist the highest.

This observation was repeated for the RMSD in acceleration (Table 3 and Figure 2). The linear movement trials had the closest agreement between methods for velocity. The 180⁰ change of direction movement trial resulted in the highest RMSD values for acceleration.

Table 2. RMSD velocity (m·s^-1) between Catapult devices and Vicon motion analysis system.

		N	Mean	Median	IQR
	All	308	0.19	0.14	0.11 – 0.21
Device location	Vest	79	0.14	0.12	0.09 – 0.15
	Back	77	0.20	0.15	0.12 – 0.22
	Front	76	0.24	0.17	0.14 – 0.27
	Left Side	76	0.18	0.13	0.11 – 0.18
Movement type	Linear	140	0.17	0.12	0.09 – 0.18
	COD 45	44	0.27	0.22	0.14 – 0.32
	COD 90	43	0.18	0.15	0.11 – 0.20
	COD 180	48	0.18	0.14	0.13 – 0.17
	Circuit	33	0.22	0.18	0.17 – 0.24

Table 3. RMSD acceleration (m·s^-2) between Catapult devices and Vicon motion analysis system.

		N	Mean	Median	IQR
	All	308	0.79	0.76	0.64 – 0.91
Device location	Vest	79	0.75	0.74	0.62 – 0.81
	Back	77	0.77	0.76	0.60 – 0.91
	Front	76	0.85	0.83	0.66 – 0.97
	Left Side	76	0.79	0.79	0.65 – 0.88
Movement type	Linear	140	0.69	0.73	0.60 – 0.81
	COD 45	44	0.67	0.67	0.48 – 0.87
	COD 90	43	0.75	0.74	0.66 – 0.85
	COD 180	48	1.24	1.25	1.16 – 1.32
	Circuit	33	0.77	0.80	0.73 – 0.86

Figure 2. RMSD velocity (first row) and acceleration (second row) results for each trial, stratified by device location (first columns) and movement type (second column).

Table 4 shows the MAD in the sample-to-sample distance derived from Vector T7 and Vicon positional tracking data. Across all trials the mean difference was 0.39 m, differences were larger in the device worn at the front of the waist, and during the circuit movement trial.

Table 4. MAD in sample-to-sample distance (m) between Catapult devices and Vicon motion analysis system.

		N	Mean	Median	IQR
	All	308	0.04	0.03	0.02 – 0.05
Device location	Vest	79	0.03	0.03	0.02 – 0.04
	Back	77	0.04	0.03	0.02 – 0.04
	Front	76	0.05	0.04	0.03 – 0.06
	Left Side	76	0.04	0.03	0.03 – 0.04
Movement type	Linear	140	0.03	0.03	0.02 – 0.04
	COD 45	44	0.04	0.04	0.03 – 0.05
	COD 90	43	0.04	0.03	0.03 – 0.04
	COD 180	48	0.04	0.03	0.03 – 0.04
	Circuit	33	0.07	0.06	0.05 – 0.09

Static trials

The median and IQR sample-to-sample distance for all stationary devices was zero (Table 5), indicating that in the majority of timesteps the devices did not change their x or y coordinates.

However, devices did record some changes in position such that the mean distance between samples was approximately 1-2 mm. Over 10 minutes there was no strong directional bias, and the final location of devices was very close to their initial positions (final displacement on the order of a few centimetres).

The proprietary data processing and filtering included in the manufacturer’s calculation of cumulative distance (the ‘odometer’ variable) was able to correct for the small changes in position and returned total distances of less than 0.02 m.

Table 5. Results of static trials (three devices left stationary for 10 minutes).

	Derived from x, y positions
	Distance between successive samples (m)			Displacement at final sample (m)	Odometer at final sample (m)
Device	Mean	Median	IQR
All	0.001898	0	0 – 0
1	0.001644	0	0 – 0	(-0.02, -0.05)	0.02
2	0.001548	0	0 – 0	(-0.03, -0.06)	0.00
3	0.002502	0	0 – 0	(-0.06, -0.05)	0.01

→ Download this Vector T7 White Paper by clicking here.

5. SUMMARY

• Vector T7 devices had a low RMSD for measures of velocity and acceleration and a low MAD for measures of distance during movement trials involving high rates of acceleration, deceleration and change of direction.
• Vector T7 devices displayed a similar low RMSD for measures of acceleration across all movement trials with the exception of a 180-degree change-of-direction where RMSD was slightly higher. This is likely due to the movements in this trial involving higher rates of acceleration/deceleration and suggests error increases as the rate of change in velocity increase. However, this error can still be considered low (mean RMSD of 1.24). 
• Vest placement resulted in the lowest error for measures of velocity, acceleration, and distance while placement of the device at the front of the waist resulted in the highest error. Regardless of device placement all locations resulted in low error across all measures with all RMSD ≤ 0.85. 
• The positional stability of Vector T7 devices when stationary was high with a mean distance between samples of ~1-2 mm.
• The low error suggests the Vector T7 device used in conjunction with ClearSky provides a valid measure of velocity, acceleration, and distance during team-sport specific tasks, including linear maximal efforts and change-of-direction movements.

→ Download this Vector T7 White Paper by clicking here.

6. REFERENCES

1. Malone, J. J., Lovell, R., Varley, M. C., & Coutts, A. J. (2017). Unpacking the black box: applications and considerations for using GPS devices in sport. International journal of sports physiology and performance, 12(s2), S2-18.
2. Linke, D., Link, D., & Lames, M. (2018). Validation of electronic performance and tracking systems EPTS under field conditions. PloS one, 13(7), e0199519.
3. Linke, D., Link, D., & Lames, M. (2020). Football-specific validity of TRACAB’s optical video tracking systems. PloS one, 15(3), e0230179.
4. Luteberget, L. S., Spencer, M., & Gilgien, M. (2018). Validity of the Catapult ClearSky T6 local positioning system for team sports specific drills, in indoor conditions. Frontiers in physiology, 9, 115.
5. Serpiello, F. R., Hopkins, W. G., Barnes, S., Tavrou, J., Duthie, G. M., Aughey, R. J., & Ball, K. (2018). Validity of an ultra-wideband local positioning system to measure locomotion in indoor sports. Journal of sports sciences, 36(15), 1727-1733.
6. Hodder, R. W., Ball, K. A., & Serpiello, F. R. (2020). Criterion validity of Catapult ClearSky T6 local positioning system for measuring inter-unit distance. Sensors, 20(13), 3693.
7. R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/
8. Van Boxtel, G.J.M., Laboissière, R., & Wilhelm, H.D. (2021). gsignal: Signal processing. URL: https://github.com/gjmvanboxtel/gsignal

→ Download this Vector T7 White Paper by clicking here.