TECHNICAL REPORT

VALIDATION OF AN ACCELEROMETER MEASURING SPORT PERFORMANCE

KIMITAKE SATO,1 SARAH L. SMITH,2
1 2

FOR

AND

WILLIAM A. SANDS2

Department of Sport and Exercise Science, University of Northern Colorado, Greeley, Colorado; and Performance Services Division, United States Olympic Committee, Colorado Springs, Colorado

ABSTRACT
Sato, K, Smith, SL and Sands, WA. Validation of an accelerometer for measuring sport performance. J Strength Cond Res 23(1): 341347, 2009Weightlifting technique is a well-studied subject with regard to standard biomechanical analysis that includes barbell velocity as well as barbell trajectory, but kinematic data such as barbell acceleration have not often been reported. Real-time or near-real-time feedback can be more helpful to coaches and athletes than delayed feedback. The purpose of this study was to validate measures obtained by a commercially available accelerometer in comparison with kinematic data derived from video. The hypothesis was that there would be a high positive relationship between accelerometer data and acceleration measures derived from video records of a barbell high-pull movement. Accelerometer values and kinematic data from high-speed video were obtained from 7 volunteers performing 2 trials each of a barbell high-pull. The results showed that the accelerometer measures were highly correlated with derived acceleration data from video (r = 0.940.99). On the basis of these results, the device was considered to be validated; thus, the unit may be a useful tool to measure acceleration during real-time training sessions rather than only reserved for collecting data in a laboratory setting. This device can be a valuable tool to provide instant feedback to coaches and athletes to assess individual barbell acceleration performance.

KEY WORDS barbell acceleration, barbell movement, resistance

training

INTRODUCTION

iomechanical characteristics of Olympic lifts (snatch and clean & jerk) have been well documented in the literature in terms of describing lifting technique (211,13,14). In the majority of studies, barbell velocity and barbell trajectory have been
Address correspondence to Kimitake Sato, jpnsatok@hotmail.com. 23(1)/341347 Journal of Strength and Conditioning Research 2009 National Strength and Conditioning Association

analyzed as well as mechanical work and power output (2,3,513). However, Stone et al. (13) described difculty in predicting successful from unsuccessful lifts on the basis of these variables. Barbell acceleration analysis in weightlifting is limited to 2 previously published studies (6,7). In both investigations, barbell acceleration data were graphically displayed; however, there was no discussion regarding the usefulness and application of vertical barbell acceleration in either interpreting lifting performance or as a means for improving technique. Further analysis is necessary to understand the following factors: a) vertical acceleration patterns in weightlifting, b) effect of acceleration rate on individual technique during the lifts, c) possible acceleration changes over a single training session, and d) possible acceleration changes over a long-term training schedule. Unlike barbell velocity measures, acceleration values directly relate to force production generated by a lifter with various weight loads. Additionally, measuring barbell acceleration at a higher sampling frequency than was reported in past studies (6,7) may make it possible to detect small but relevant changes in various phases of weightlifting. If an accelerometer can be attached to the barbell with minimal disturbance of the ow of the training session, the instrument may be able to provide an indication of fatigue or overtraining. Therefore, attaching a compact, lightweight device without compromising the weightlifters technique to measure acceleration may show promise as a valuable feedback tool. If the instrument is also relatively inexpensive, it can be made available to more lifters. The measurement of training volume (sets, repetitions, and weights) in resistance training has been used as an indicator for coaches in estimating strength improvement or analyzing the development of fatigue (1,4,14,15). The continued use of such a methodology may be one of the most reliable and accurate ways to train weightlifters as well as other athletes incorporating weightlifting in their resistance training program. However, recent biomechanical studies have shown great interest in measuring kinetics (force production measured by force plates) as well as kinematics (barbell trajectory and velocity) of weightlifting to analyze force and movement patterns (59,1113). Barbell trajectory analysis displays possible asymmetric barbell movement patterns (3,59,11,13), whereas barbell velocity analysis displays speed
VOLUME 23 | NUMBER 1 | JANUARY 2009 |

341

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Validation of an Accelerometer
of the lifts (5,7,8,11,13). Instant feedback is desirable for coaches and athletes to evaluate athletic performance. Moreover, to minimize the risk of overtraining, it may be necessary to understand the acceleration pattern of each lift during higher workloads of the training period to detect anomalies that merit attention. Acceleration data for each phase of weightlifting have been presented but not well explained, and these data need further investigation (6,7). The purpose of this study was to validate acceleration data collected by an accelerometer with similar data obtained from a kinematic analysis of a specic movement sequence recorded by high-speed video. The tested hypothesis was that there was a strong positive relationship between accelerometer data and derived acceleration data. One important limitation of the accelerometer became apparent. The unit is directionally dependent and must remain in a xed position throughout the motion being investigated to measure vertical acceleration accurately; rotation of the device causes miscalculation of vertical acceleration. During a pilot study, the accelerometer seemed to remain in vertical alignment during most stages of the lift, but some rotational movement was obvious during the catch phase. Given the rotation of the accelerometer while attached to a barbell during Olympic lifts, the high-pull motion was selected as the requisite movement for the validation study because the motion does not cause rotation of the barbell. unit is to record forward-backward, up-down, and sideto-side acceleration in the X, Y, and Z directions, respectively. The accelerometer weighs 56.7 g and is 3.8 cm in width, 8.8 cm in length, and 3 cm in depth (see Figure 1). The accelerometer is attached to a Bluetooth wireless device (PASCO Pasport Airlink SI). The wireless device is also lightweight and compact, weighing 113.4 g and measuring 4 3 13 3 2.5 cm, with the length measure including the antenna. Total weight of the device is 170.1 g; maximum size is 4 3 20.2 3 3 cm. Total length of the 2 units is reduced when both units are mated (see Figure 2). The total weight of the unit is equivalent to a plastic barbell collar that typically holds weight plates in place. The accelerometer unit was securely attached by 2 hose clamps to a plastic barbell collar; placement of the sensor unit was on the right end of the barbell in relation to the position of

METHODS
Experimental Approach to the Problem

An easily administered testing device is useful for coaches to assess athletes conditions. In this study, a commercial accelerometer was tested to validate the accuracy of barbell acceleration data with kinematic data obtained from highspeed video. Correlation analysis was used to measure the relationship.
Subjects

Figure 1. Accelerometer.

Seven men volunteered for this study (age: 24.29 6 2.98 years old; height: 175.62 6 6.95 cm; mass: 82.91 6 12.91 kg). They had no injuries at the time of data collection. Three subjects were competitive weightlifters who perform high-pulls and other explosive exercises on a daily basis. The other 4 subjects were not competitive, but they were experienced in resistance training. Data were collected in compliance with policies of the United States Olympic Committee on the testing of athletic subjects.
Instrumentation

A 3-axis accelerometer (PS-2119, PASCO, Roseville, Calif ) measured acceleration in the X, Y, and Z planes. Although the 3 sensors are orientation dependent, it is possible to obtain an accurate resultant acceleration if certain software features are disabled. Orientation of the sensor must remain constant throughout the movement sequence to avoid misrepresentation of vertical acceleration. Standard conguration for this

Figure 2. Accelerometer is attached to Bluetooth transmitter.

342

Journal of Strength and Conditioning Research

the

TM

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Journal of Strength and Conditioning Research

the

TM

| www.nsca-jscr.org

Figure 3. Accelerometer and transmitter attachment to the barbell.

Calif ), which acquires, displays, and analyzes the data. The acceleration curve is displayed in real time. A high-speed video camera (HSV-400, NAC Image Technology, Japan) was positioned at a distance of 4.8 m from the end of the barbell on the right side of the lifter. The camera was perpendicular to the plane of barbell movement. A reective marker 3.175 cm in diameter was placed on the right end of the barbell for automatic digitizing of the barbell trajectory (see Figure 5). Sampling occurred at 100 Hz. Thus, there was a corresponding match between each video frame depicting barbell position and each acceleration data value. The high-speed analog images were recorded onto VHS tape and subsequently copied to miniDV tape through a Canopus (ADVC100, San Jose, Calif ) analog-to-digital video converter. All movement sequences were then captured by Peak Motus (Vicon-Peak, version. 8.5, Centennial, Colo); barbell X and Y coordinate position data were captured through the automatic digitizing feature. A Butterworth lter with a cutoff frequency of 4 Hz was used to smooth the video data. Vertical barbell displacement, velocity, and acceleration data for each of the 14 movement sequences were calculated.
Procedures

Figure 4. Measurement: vertical direction.

All subjects reported to the Athlete Performance Laboratory at the Colorado Springs Olympic Training Center (CSOTC, Colorado Springs, Colo). After performing self-selected stretching and warm-up exercises, they were instructed to perform a barbell high-pull with a 10-kg plate on each end of the barbell; total weight lifted was 40 kg. Each participant performed 2 trials. Video images of each barbell movement were recorded from before the initiation of barbell liftoff from the platform (Figure 6) until just after the highest trajectory point (Figure 7) during the high-pull by the high-speed video system; simultaneously, acceleration data were transmitted by Bluetooth technology to a laptop computer.
Statistical Analyses

the lifter (see Figure 3). Because the sensor unit was attached underneath and in line with the long axis of the barbell, the Z-axis channel of the unit collected acceleration data in the vertical direction (see Figure 4). A sampling rate of 100 Hz was selected for data collection, which is the highest data collection rate the device can provide. During data collection, acceleration data are transmitted to a laptop computer running DataStudio software (PASCO Scientic, Roseville,

Both raw and ltered accelerometer data were graphically displayed to compare and contrast the 2 datasets for each of

Figure 5. Data collection layout.

VOLUME 23 | NUMBER 1 | JANUARY 2009 |

343

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Validation of an Accelerometer

Figure 6. Start position of high-pull.

Figure 7. End position of high-pull.

the 14 trials. Although visual inspection of the graphs revealed no discernible differences, a Pearson product-moment correlation coefcient was calculated to determine the specic relationship within each trial for the raw and ltered accelerometer data (Microsoft Ofce Excel 2003, Redmond, Wash). For all 14 trials, the derived acceleration data were then graphically displayed with both raw and ltered accelerometer data. Again, visual inspection showed close agreement between all 3 (raw acceleration, smoothed/ltered acceleration from the accelerometer, and smoothed/ltered acceleration from video) datasets. Correlation analyses were performed to determine the relationship between the derived acceleration data with the raw accelerometer data.

data is depicted in Figure 8. Although this graph represents data for only one trial for one subject, a similar pattern was observed for all other trials. The comparison between accelerometer raw data and ltered data showed a correlation coefcient of r = 0.99 for all 14 trials (see Table 1). On the basis of these results, raw

RESULTS
An example of the results of the graphic comparison between raw and ltered accelerometer
Figure 8. A graphic example of acceleration match-up data between raw data and ltered data.

344

Journal of Strength and Conditioning Research

the

TM

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Journal of Strength and Conditioning Research

the

TM

| www.nsca-jscr.org

TABLE 1. Pearson product-moment correlation coefcients between raw acceleration data and ltered acceleration data sets. Subjects 1 2 3 4 5 6 7 High-pull trial 1 r r r r r r r = = = = = = = 0.99 0.99 0.99 0.99 0.99 0.99 0.99 High-pull trial 2 r r r r r r r = = = = = = = 0.99 0.99 0.99 0.99 0.99 0.99 0.99

Correlation coefcient values ranged from r = 0.94 to 0.99 for all subjects across 2 trials (see Table 2). All relationships between raw accelerometer data and the corresponding acceleration data obtained from the kinematic analysis were statistically signicant (r = 0.87, p , 0.01). Because the level of weightlifting experience among participants was varied, the acceleration proles were considered to be dependent on each subjects strength and technique.

DISCUSSION
The objective of this study was to validate data collected from an accelerometer with acceleration data derived from a kinematic analysis of a movement sequence recorded by high-speed video. The results show that the accelerometer displayed nearly identical acceleration values compared with the derived acceleration values and support the hypothesis that a strong positive relationship existed between the 2 datasets. This inexpensive, commercially available accelerometer has demonstrated the capability to collect valid barbell acceleration values during a high-pull movement sequence. The accelerometer was shown to be an instrumentation device that could easily be attached to a barbell. The required laptop computer running the acquisition software was easily positioned so as to provide little if any distraction to the subjects during their lifting trials. Thus, neither subjects nor their equipment were encumbered by cables attaching sensors to a data acquisition unit or laptop computer.

datasets from the accelerometer were then selected for comparison with the acceleration data obtained from the kinematic analysis of barbell movement captured by highspeed video. A typical graphic comparison between raw and ltered accelerometer data and the corresponding acceleration data obtained in the kinematic analysis is presented in Figure 9. In all of the other trials, graphs showed the corresponding acceleration datasets displaying a similar curve. Results of the correlation analyses show a high positive relationship between the accelerometer raw data and the derived acceleration data from kinematic analysis.

Figure 9. A graphic example of acceleration match-up data between raw data and kinematic analysis obtained from high-speed video.

VOLUME 23 | NUMBER 1 | JANUARY 2009 |

345

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Validation of an Accelerometer
within approximately 30 m; thus, distance between the accelerometer and the receiver is not thought to be a serious constraint. The second limitation is the durability of the accelerometer. Ultimately, the accelerometer needs to provide meaningful, relevant, and accurate information that impacts performance enhancement while being used within the connes of a realistic weightlifting training session. Initially, the need for validating this unit was for the expressed purpose of being able to measure barbell acceleration patterns during Olympic lifts. However, the accelerometer was not designed or developed for the purpose of absorbing the shock of being dropped onto a lifting platform from the height of a barbell raised overhead by a weightlifter. Excessive external shock not only interferes with the acceleration reading, but, in certain situations, it may also destroy the device. If there is a possibility of external shock to the accelerometer, an adequate amount of protection is needed to ensure its viability. This need eventually led to the design of the mounting foam unit shown in Figure 10. Subsequent pilot testing demonstrated that this padded mount provided sufcient protection for the accelerometer. The nal and most relevant limitation was that the individual (X, Y, and Z) acceleration data were directionally dependent. As stated earlier, this property of the accelerometer unit prevented the collection and analysis of barbell vertical acceleration data during the execution of snatch lifts. Because rotation of the bar in the catch phase caused the accelerometer to change its position from the initial vertical orientation, it became necessary to use the high-pull motion for validating the vertical acceleration pattern. As previously mentioned, the collection of barbell acceleration data and subsequent assessment have not been well studied (6,7). Because acceleration patterns directly mirror force production, they may be useful in understanding

TABLE 2. Pearson product-moment correlation coefcients between raw acceleration data and acceleration data obtained from kinematic analysis. Subjects 1 2 3 4 5 6 7 High-pull trial 1 r r r r r r r = = = = = = = 0.98 0.99 0.99 0.98 0.97 0.95 0.97 High-pull trial 2 r r r r r r r = = = = = = = 0.94 0.99 0.99 0.98 0.99 0.99 0.96

Additionally, the sensors and software required a minimum time for setup in data collection sessions. Most importantly, the instrument provided immediate results and feedback of movement performance. Current literature in weightlifting describes key components of weightlifting from both the kinematic and kinetic perspectives (2,3,514). Although past studies have identied biomechanical characteristics of weightlifting techniques, none of the data were collected in real time during training sessions (3,5,6,711,13). In additional pilot work, the accelerometer was actually attached to the barbell of several weightlifters during training sessions to identify the barbell vertical acceleration patterns of their Olympic lifts with various workloads. The data were ready to display immediately after each trial for quick feedback to coaches and athletes; data collected after the catch phase were ignored. During pilot sessions, acceleration patterns for various athletes were shown to change during the course of a training session, with patterns generally reecting lower acceleration values. This decrease in barbell acceleration may be related to fatigue. Instead of using the method of deciding on the training volume before the particular training session, this instrument may lead to determination of an optimal range of training loads for the individual lifters during the actual training session. Training load could be based on the performance that the athlete displays for any given day on a lift-to-lift basis. Only one study was found that analyzed fatigue trends in weightlifting, but the data were collected in a laboratory setting (7). The accelerometer may ultimately be found to be a simpler, more economical way to collect relevant data about workloads that can be used in a realworld weightlifting training venue. There are some limitations in using the accelerometer. First, there is the need for line-of-sight between the sensor and receiver for wireless operation. At the time of data collection, distance between the accelerometer and the laptop was approximately 3 m. This distance was similar during pilot work in the weightlifting room. The connection was not lost

Figure 10. Mounting procedure for barbell acceleration analysis.

346

Journal of Strength and Conditioning Research

the

TM

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.

Journal of Strength and Conditioning Research

the differences in lifting technique that occur not only within a training session but also over multiple training sessions. The possibility of determining the number of sets, repetitions, and workload for a training session based on information directly related to performance and fatigue/nonfatigue status needs to be explored. In summary, the use of an accelerometer that is easily attached to a barbell and that provides real-time results needs further study to determine its usefulness.

the

TM

| www.nsca-jscr.org

improve their performances on the basis of scientic information rather than on time-consuming trial and error.

REFERENCES
1. Baechle, TR and Earle, RW. Essentials of Strength Training and Conditioning (2nd ed.). Champaign: Human Kinetics, 2000. pp. 412415. 2. Barton, J. Are there general rules in snatch kinematics? In: Proceedings of the Weightlifting Symposium. Olympia: International Weightlifting Federation, 1997. pp. 119128. 3. Bartonietz, KE. Biomechanics of the snatch: toward a higher training efciency. Strength Cond J 18(3): 2431, 1996. 4. Garhammer, J and Hateld, FG. Weightlifting. In: Encyclopedia of Physical Education, Fitness, and Sport. Cureton, TK, ed. Reston: AAHPERD, 1985. pp. 595606. 5. Gourgoulis, V, Aggeloussis, N, Kalivas, V, Antoniou, P, and Mavromatis, G. Snatch lift kinematics and bar energetics in male adolescent and adult weightlifters. J Sports Med Phys Fitness 44: 126131, 2004. 6. Gourgoulis, V, Aggeloussis, N, Mavromatis, G, and Garas, A. Threedimentional kinematic analysis of the snatch of elite Greek weightlifters. J Sports Sci 18: 643652, 2000. 7. Haff, GG, Whiteley, A, McCoy, LB, OBryant, HS, Kilgore, JL, Haff, EE, Pierce, K, and Stone, MH. Effects of different set congurations on barbell velocity and displacement during a clean pull. J Strength Cond Res 17: 95103, 2003. 8. Hiskia, G. Biomechanical analysis on performance of world and Olympic champion weight lifters. In: Proceedings of IWF Coaching, Referring & Medical Symposium. Olympia: International Weightlifting Federation, 1997. pp. 137158. 9. Isaka, T, Okada, J, and Funato, K. Kinematics analysis on the barbell during the snatch movement of elite Asian weight lifters. J Appl Biomech 12: 508516, 1996. 10. Lee, YH, Huwang, CY, and Tsuang, YH. Biomechanical characteristics of preactivation and pulling phases on snatch lift. J Appl Biomech 11: 288298, 1996. 11. Schilling, BK, Stone, MH, OBryant, HS, Fry, AC, Coglianese, RH, and Pierce, K. Snatch technique of collegiate national level weightlifters. J Strength Cond Res 16: 551554, 2002. 12. Souza, AK, Shimada, SD, and Koontz, A. Ground reaction forces during the power clean. J Strength Cond Res 16: 423427, 2002. 13. Stone, MH, OBryant, HS, Williams, FE, Pierce, K, and Johnson, RL. Analysis of bar paths during the snatch in elite male weightlifters. Strength Cond J 20(4): 3038, 1998. 14. Stone, MH, Pierce, K, Sands, WA, and Stone, ME. Weightlifting: a brief overview. Strength Cond J 28(1): 5066, 2006. 15. Stone, MH, Pierce, K, Sands, WA, and Stone, ME. Weightlifting: program design. Strength Cond J 28(2): 1017, 2006.

PRACTICAL APPLICATIONS
Because of its capability for capturing data and real-time display, this commercial accelerometer could be a valuable tool for eld application. Because the unit collects data that reect force production for lifting during a training session, is easily attached, leaves the athlete unrestrained by cables, and is not cost-prohibitive, it seems worthwhile to study its potential for improving lifting performance. By quantifying barbell acceleration during training sessions, an accelerometer was considered a valuable tool during our work with athletes and coaches. When there was a change or decline in barbell acceleration during the high-workload sessions, the pattern was immediately observable; the information gained gave both the coach and lifter the opportunity to consider whether the workload needed to be adjusted or terminated. Use of an accelerometer as a fatigue indicator may be helpful in detecting the early signs of fatigue. The ndings of Haff et al. (7) show that an adequate amount of rest between repetitions prevented decreases in barbell velocity when compared with a condition of no rest between lifts. Because it is important to have an adequate amount of rest between sets to fully recover, any disparity within the measured acceleration patterns for the total number of repetitions within a set was clearly shown. Whether within a single workout or over multiple training sessions, knowledge of changes in lifting acceleration patterns can be of benet to athletes and coaches. The potential for quantifying workout volume comes closer to being a reality. In summary, wherever force production is required to initiate any sport motion, acceleration patterns and values are critical to describing and understanding the movement. By measuring this aspect of the sporting activity, athletes may

VOLUME 23 | NUMBER 1 | JANUARY 2009 |

347

Copyright N ational S trength and Conditioning A ssociation. Unauthorized reproduction of this article is prohibited.