Advertisement

Setting up the Experimental Design

  • Gonçalo DiasEmail author
  • Micael S. Couceiro
Chapter
  • 734 Downloads
Part of the SpringerBriefs in Applied Sciences and Technology book series (BRIEFSAPPLSCIENCES)

Abstract

The past decades have seen technological advances and new analysis tools that allow comprehensive interpretation of human movement patterns. As movement patterns are being analyzed with improved technology, subtle individualities or signature patterns of movement have been identified. These technological advances open a window of opportunity to further study and augment understanding in several fields, including sport science, that benefit from a multidisciplinary approach (biomechanics, engineering, mathematics, motor control) (Dias et al. On a ball’s trajectory model for putting’s evaluation. Computational intelligence and decision making—trends and applications, from intelligent systems, control and automation: science and engineering bookseries 2013). This chapter may be seen as a guide for setting up the experimental design for research purposes. Several methodological and technological alternatives will be presented and compared. Moreover, a case study provided by the authors and applied to golf putting in the laboratory context will be introduced as an example. Note that the same case study will be used throughout the book. New methods for performance analysis of golf putting have been suggested, focusing on individual kinematic analysis, rather than the traditional pooling of group data, such as in Couceiro et al. (J Motor Behav 45:37–53, 2013) and Dias et al. (On a ball’s trajectory model for putting’s evaluation. Computational intelligence and decision making—trends and applications, from intelligent systems, control and automation: science and engineering bookseries 2013). The purpose was to understand the relevant changes resulting from the interaction between the athlete’s characteristics and the surrounding context by analyzing the motor behavior profiles as measured by the individual kinematic strategies. It is also important to mention that the use of new technologies involves a more profound approach than the traditional linear techniques, which mainly consider product measurements/variables that are a result of the movement (through average, standard deviation or variation coefficient), something that does not allow analysis of the movement itself (Harbourne and Stergiou JNPT Am Phys Ther 89:267–82, 2009).

Keywords

Golf putting Task Procedures Instrumentation Performance 

3.1 Setup

This section explains how one should proceed to prepare and build an experimental setup to analyze golf putting. Note that although we will focus on the golf putting case study, many of the choices presented here may easily be put into practice for other pendulum-like motion and certain ballistic-like actions, such as the tennis serve. This section considers the insights from previous publications, such as Couceiro et al. (2013) and Dias et al. (2013), refining them to a more detailed and tutorial description, and indicating how one can choose the sample (participants), define the task, the procedures and the relevant variables.

3.1.1 Participants

The choice of the participants is of the utmost importance as there is a wide difference between motor execution from novices (who are still learning how to perform a given movement) and from experts (which already completely dominate the movement). When preparing a scientific study, the aim is to ensure the replicability of the obtained results. In spite of this, one must establish the object of analysis and clearly define the criteria considered for sample selection. One of the most important criteria is the level of a player’s expertise which will have an influence on dependent and independent variables that will be handled in the research. For instance, Couceiro et al. (2013) analyzed the performance of golf players by comparing their putting patterns through classification methods. By analyzing that work, or any other related work regarding golf putting or any other action, one can easily conclude that novice players do not usually present any sort of playing patterns, i.e., their actions are more chaotic, even random, and present a higher level of variability. Bearing this idea in mind, the authors chose a sample of expert players with a handicap <15, with >10 years of golf practice and participation in national golf tournaments. This is the criteria that any other research should meet to confirm the results obtained by Couceiro et al. (2013).

Besides the level of expertise, other requirements may also influence the final outcome of the research. Namely, in the same study, players were aged 32 ± 10 years, volunteers, male, and right-handed (Couceiro et al. 2013). Although age may not be a critical factor on golf putting performance, researchers will still have to state the average value and standard deviation to improve the level of repeatability of the results. To be even more demanding in the description of the sample, all participants should have legal capacity and competence to participate voluntarily in the investigation. Thus, all athletes are informed in advance that they are participating freely in the study and that they also have complete freedom to withdraw at any time without any penalty. Furthermore, in similar studies, one should state that the study is conducted in accordance with the code of ethics of the proposing institution and the recommendations of the Declaration of Helsinki on human research. 1 In other words, performers should not have suffered from any physical or mental disability, being previously informed by written consent that the research does not cause any kind of damage to their physical and mental integrity.

3.1.2 Task and Procedures

Besides describing the sample, the characteristics of the task, the structure of the motor task and the experimental procedures considered should be thoroughly described in the scientific research.

Defining and manipulating the characteristics of the task is important to obtain consistent results with the existing state of the art, while at the same time, it should offer a realistic and challenging context to players. For instance, a very simple task may discourage an expert athlete who already has experience in golf putting, thus resulting in a performance that does not describe his/her real capabilities. On the other hand, a very demanding task may have the opposite effect and discourage an inexperienced athlete who fails to achieve the desired objectives of the researcher. Although obvious, the misconception of the task is very common in scientific research and contaminates the entire study (Guadagnoli and Lee 2004).

The same can be said about the general structure of the motor task. However, in general, this topic has been thoroughly studied by several researchers within the field of motor control (Tani 2005; Davids et al. 2008; Magill 2011), in which particular attention was given to the contextual interference effect. This phenomenon is related to the degree of interference that occurs from the learning and performance of a motor skill. Within the several studies on the contextual interference effect in golf putting, such as Tani (2005) and Magill (2011), researchers have identified three possible ways to organize and handle the task (Table 3.1).
Table 3.1

Types of motor practice in learning golf putting

Types of motor practice

Distance to the hole: 1, 2 and 3 m

Blocks

1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3

Series

1,2,3, 1,2,3, 1,2,3, 1,2,3

Random

1,3,2, 3,2,3, 2,1,3, 2,3,2

The theoretical assumptions that support the contextual interference effect point toward better results in the acquisition phase using a block practice over the alternatives (for in experienced golfers). In contrast, whenever the idea is to retain or transfer the learning, a random practice seems to result in higher levels of motor performance (expert golfers) (Battig 1966; Shea and Morgan 1979; Dias and Mendes 2010).

However, authors such as Porter and Magill (2005) or Dias and Mendes (2010) argue that it may be beneficial to promote a gradual increase of the contextual interference in both learning and evaluation of putting, in order to achieve positive effects, especially in the transfer phase, where it is required to produce a movement similar to the previous one, but with different changes in their temporal characteristics or intensity, such as hitting the ball in a curvilinear trajectory towards the hole.

This hybrid organization means that the practice sessions begin with structured blocks, being followed by series and subsequently by random practice conditions. Nevertheless, any of these studies quantitatively confirm this assumption, thus making it, so far, unequivocally inconclusive and worthy of further exploitation.

Adducing further contributions to these theoretical assumptions, and although it is not considered as a classical hypothesis to explain the contextual interference effect, the ‘hypothesis’ that was formulated by Guadagnoli and Lee (2004), denoted as challenge point, allows better understanding of the influence of having different levels of complexity in the learning process of motor skills. Under this premise, one can speculate that it is necessary to create an optimal learning/adapting level, such that an inexperienced golfer can adapt to the complexity of golf putting. As previously stated, a low level of complexity may not result in significant contextual interference effects, as it may not motivate the athlete. In contrast, a high level of complexity hardly results in high levels of motor proficiency, as it may discourage the athlete to persist (Guadagnoli and Lee 2004).

Following this idea, we note that authors such as Bjork (1994, 1999), argue that increasingly gradual contextual interference can promote an efficient outcome and stabilize the levels of this effect. Nevertheless, it is concluded that the characteristics of the task and the level of expertise of the athlete are both variables that one needs to consider when designing practice for motor learning in the context of golf putting (Dias and Mendes 2010). For instance, in a previous study by Dias et al. (2014), the main objective was to investigate the adaptation to external constraints and the effects of variability in a golf putting task. The results show that the players changed some parameters to adjust to the task constraints (slope and putting distance), namely the duration of the backswing phase, the speed of the club head and the acceleration at the moment of impact with the ball. Hence, the effects of different golf putting distances in the no-slope condition on different kinematic variables suggest a linear adjustment to distance variation that was not observed during the slope condition. Moreover, the data also indicate that the speed of impact on the ball (process variable) is the one showing a stronger correlation with magnitude of the radial error (product variable), making that variable the best single predictor of golf putting performance.

3.1.3 Variables

Research variables are very important to analyze the sample under a given domain. Moreover, they also determine the type of instruments, organization and experimental procedures that are adopted. As we have seen earlier, it is necessary to know the state of the art in that domain and understand the most typical product variables related to the final result of the operation (e.g., the number of balls that enter the hole), as well as the most typical process variables related to the player’s performance (e.g., duration of the movement). During the conception of the experimental setup, one needs to go further and contextualize the independent and dependent variables under consideration.

Independent variables are intentionally manipulated by the researcher and may assume several states throughout the experiments. In the analysis of golf putting, the independent variables can match the type of structure of the motor task, which may be translated into different types of induced constraints, such as the distance to the hole (e.g., striking the ball at 1, 2, 3 and 4 m to the hole in a straight path) or the addition of a slope between the trajectory of the ball and the hole.

Dependent variables are those that the researcher intends to assess. Typically, the research focuses on understanding how dependent variables are affected by the independent variables. In a study involving golf putting, the dependent variables are related with the performance of the athlete, not only in terms of product, but also in terms of process. They may, for instance, include the distance between the final position of the ball and the hole (radial error), as well as the duration, amplitude, maximum velocity and maximum acceleration of each phase (backswing, downswing, ball impact and follow-through).

3.2 Instrumentation

The instruments used to analyze a given sport movement need to be adapted to the purpose of the research as well as the movement. For instance, while high speed cameras (>200 Hz) may be necessary for a ballistic-like motion, such as the tennis serve, golf putting may be analyzed with a much lower frame rate (e.g., 30 Hz). Nevertheless, and depending on the level of detail one may need (e.g., analyze the tilt of the putter head), high resolution cameras may still be required.

Cameras or any other instruments (e.g., accelerometers) may provide a detailed feedback about the motion and the final result of the action that may not be ensured, at least accurately, by the naked eye.

3.2.1 Traditional Cameras

In golf putting, as with most pendulum-like movements, most of the analysis can be achieved with the use of bidimensional tracking systems, such as a single traditional camera, e.g., complementary metal oxide—semiconductor cameras. This simple method allows one to assess valuable and accurate information about the putting action or ball trajectory which, on its own, can provide important feedback to athletes and coaches (Neal and Wilson 1985; Couceiro et al. 2012). New ever-improving and low-cost cameras have allowed successive important breakthroughs in most sports, including golf putting, by providing more information (high rates and high resolution) about action patterns and how these evolve over time (Neal et al. 2007; Dias et al. 2013).

Dias et al. (2013) used a photography camera (Casio Exilim/High Speed EX-FH25) featuring filming capabilities of up to 210 Hz at a resolution of 480 × 360 pixels, with a lens of 26 mm focal length. One camera was used to capture the movement, while another one, exactly the same, was used to capture the ball’s trajectory (see camera 2). Figure 3.1 illustrates the experimental setup adopted by the authors. This kind of illustration is helpful so other researchers may replicate the study. Furthermore, it provides an overall idea of the setup.
Fig. 3.1

Top and upper view of the experimental setup from Dias et al. (2013)

The main disadvantage regarding traditional cameras is the use of computationally complex, and sometimes unreliable, detection and estimation methods. For instance, Couceiro et al. (2012) presented a strategy of detection, estimation and classification applied to the golf putting context. The detection method consisted of a simple geometry and color match. Afterwards, the estimation method consisted of fitting the trajectory of the object (golf putter) with a mathematical model, in which its parameters were optimized comparing several techniques in terms of computational complexity and memory. Five different estimation techniques were studied, applied and compared, namely gradient descent, pattern search, downhill simplex, particle swarm optimization (PSO) and Darwinian particle swarm optimization (DPSO). Results confirmed the superior performance of the DPSO method (Couceiro et al. 2012).

Afterwards, several classification algorithms were considered to extract the unique signatures of each player by considering the parameters previously optimized. The work tested several classifiers, from the most traditional methods, such as the linear discriminant analysis (LDA) or the quadratic discriminant analysis (QDA), to state-of-the-art approaches, such as naive Bayes (NB) and least-squares support vector machines (LS-SVM). The classification methods were compared through analysis of the confusion matrix and the area under the receiver operating characteristic curve (AUC), and the SVM presented an overall better performance.

As one may assume, the simplicity, and low-cost of using a single monocular camera is shortly transformed into a series of other complications that most sport scientists try to avoid using, for instance, semi-autonomous or completely manual tracking software, e.g., AnaMov and Tacto. Even the most state-of-the-art computer vision methods tend to fail in unstructured or outdoor scenarios, due to occlusions, light variations, and other phenomena.

Although many ever-improving strategies have been proposed over the past few years to tackle such issues, monocular cameras are still more reliable in the laboratory workspace where one can benefit from several landmarks previously employed in the environment. To overcome these limitations, computer vision methods have been crossing the bridge from two-dimensional (2D) to three-dimensional (3D) by benefiting from depth cameras.

3.2.2 3D Depth Cameras

The development of depth cameras for pose estimation brought new opportunities for human motion analysis (Shotton et al. 2013). Among the multiple depth camera choices, stereo vision and time-of-flight technologies are perhaps the most widely used. The success of depth cameras has been successively attested by commercial systems that estimate full body poses for computer games, hand poses for action interfaces, or capture detailed head motions for facial animation (Ye et al. 2013).

The revolution of depth cameras into research started with the appearance of the Microsoft Kinect. 2 Before it came onto the market, depth imaging was costly (>1,000 Euros). However, Kinect was released for the game industry at a cost of approximately 250 Euros. Such low-cost property soon led engineers and roboticists to adapt its features to other domains, namely for simultaneous localization and mapping in robotics (Endres et al. 2012) or, more specifically in the context of this book, skeleton pose tracking in biomechanical engineering and sport sciences (Fernández-Baena et al. 2012).

Despite these achievements, computer vision techniques applied to 3D imaging require a higher computational effort than their 2D counterparts. For instance, most state-of-the-art depth computer vision techniques are able to track >48 DOF of the human skeleton up to 30 Hz. Considering that a camera of approximately the same cost can acquire approximately 200 Hz, this presents a drawback that one should consider during the design phase of the experimental setup. For instance, although the Kinect sensor, as any other similar depth camera such as the ASUS Xtion, 3 could be easily deployed to acquire a player’s kinematics while performing golf putting, for faster actions, such as the tennis serve, it would be unfeasible. Moreover, similar to their 2D counterparts, depth cameras are also susceptible to lighting conditions, thus constraining their adequate use to controlled and preferably indoor environments.

3.2.3 Motion Capture Suits

Motion capture suits have been more and more in vogue for human movement analysis (Garafalo 2010). Notwithstanding the range of different suits, these sections only focus on inertial sensor-based systems. Any alternatives to this technology (e.g., stereophotogrammetry-based systems) fall within the limitations described in the previous section, since they depend on the use of cameras which restrict their use within the laboratory workspace.

Inertial sensor-based motion capture suits, such as the MTx units from Xsens solutions,4 benefit from multisensor fusion techniques applied to inertial measurement units (IMU), comprising of gyroscopes, accelerometers and magnetometers, so as to accurately estimate the 3D position of each joint. These suits are generally equipped with wireless transmitter units that send synchronized data to a data logger communicating with a computer. Therefore, such instrumental analysis can be adopted by athletes in order to allow them to mostly concentrate on the task, without constraints imposed by camera-based solutions (e.g., light conditions). Moreover, contrary to the previous alternatives, motion capture suits are completely portable, thus offering the possibility to be applied outdoors and in the field context.

However, all these benefits do not come without a cost. The state-of-the-art Xsens suit, being currently used not only for research but also in film and game industries, costs approximately 60,000 Euros. Although one can find cheaper alternatives on the market, such as the 3DSuit solution5 costing approximately 25,000 Euros, the cost is still significantly higher than the previous two alternatives.

3.2.4 Sport-Specific Devices

All the previous devices can be used for data retrieval in most sport modalities. Nevertheless, some sport-specific devices are also available on the market, such as Sony’s Smart Tennis Sensor 6 and IngeniariusInPutter. 7 The main reason behind the development of these instrumented devices is to maintain the ecological validity of the overall setup, without additionally constraining the athlete with unrealistic situations (e.g., laboratory setup, full-body suit, etc.).

In the golf putting context, the recently developed InPutter seems to be the one device that completely maintains the ecological validity of the setup. Although other products have been developed to study and improve the putting performance, such as the SAM PuttLab, 8 the InPutter is the only one that does not require any auxiliary hardware or that is confined to laboratory experiments. In brief, InPutter is an engineered golf putter designed for research, analysis and training purposes. By benefiting from an internal IMU sensor and wireless technology, it is able to retrieve the most relevant golf putting process variables, namely the putter’s trajectory over time, speed, duration and amplitude of each phase, as well as the impact force on the ball. The system additionally includes a heart rate monitor interface compatible with Polar transmitters.9 As InPutter does not require any camera systems, markers, or system infrastructure, and given its robustness, weight and design which is similar to other traditional golf putters, it can be used in both indoor and outdoor environments. Additionally, InPutter is an internet-connected product that automatically connects to the Ingeniarius Cloud, 10 thus allowing real-time debugging and monitoring over the internet.

The InPutter was developed by Ingeniarius, Lda., a private company, in which the authors of this book played a vital role in its development; the first author of this book, Dr. Gonçalo Dias, was the key consultant and researcher during the development of the product, having the main responsibility to maintain its ecological validity. The co-author of this book, Dr. Micael Couceiro, was the co-founder of Ingeniarius, Lda. and the supervisor of the engineering development of InPutter. Given the positive feedback from many golf players, namely from the European Pitch and Putt Champion Hugo Espirito Santo, InPutter is currently being used not only within the research context, but in real-life golf training. Table 3.2 summarizes several golf-specific devices and their features.

3.3 Practical Implications

The evolution in technology, especially over the past decade, resulted in some major advances in golf putting analysis. The feedback from this analysis, being either qualitative or quantitative, is of high importance as it provides a deeper knowledge around putting. Although the presented information revolves around golf putting, it can be optimized and used to analyze other sports. Actually, a higher degree of performance analysis may allow a completely new understanding about the individual singular properties of each player. Using this pertinent information, one can adjust the training programs to the player’s specificities.

3.4 Summary

In summary, experimental procedures are of the utmost importance in the analysis of sport movements such as golf putting. In this sense, before starting operationally to study a given action such as putting, it is very important to answer the following questions:
  1. (1)

    Where will the task be performed (laboratory context, indoors, outdoors, real competition, etc.)?

     
  2. (2)

    What will the research focus on (product and process variables, physiological variables, etc.)?

     
  3. (3)

    What information is given and how it will be provided to the participants (verbally, by video, by demonstration, or mixed information)?

     
  4. (4)

    What materials will be used and manipulated by the participants (putter and golf ball), clearly stating whether they will always use the same materials throughout the study or materials will vary according to the purpose of the task?

     
  5. (5)

    What is the full range of the trials under study (distance to the hole, pose of the player, etc.)?

     

Answering these questions will help define the necessary instrumentation hardware and software that one will need for a particular case study. Moreover, it is advisable to conduct a preliminary study to fully evaluate the defined setup and determine if the answers to the previous five questions are adequate.

For instance, the work by Dias et al. (2014) evaluated the setup for expert golfers a posteriori, in works such as Couceiro et al. (2012) and Dias et al. (2013), by considering three novice players (namely the research team). That preliminary study allowed gaps for some of the answers provided to the five questions above to be filled. For example, although the number of conditions was adequately defined, each condition comprised 30 trials, thus resulting in 270 trials for the whole study. All novice athletes, although novice and considerably different in terms of motor performance than experts, showed extreme levels of fatigue and their performance significantly dropped after approximately 200 puttings. Therefore, each condition was resized to 20 trials each, thus resulting in 180 trials for the whole study.

That preliminary study also allowed a rethink about adequate data acquisition hardware. To further improve the accurate identification of the ball impact phase, the putter was equipped with an accelerometer. Nevertheless, the use of two cameras (Fig. 3.1) and one accelerometer led to the development of a trigger mechanism so as to synchronize the instrumentation setup (Fig. 3.2).
Fig. 3.2

Experimental setup developed for data synchronization (adapted from Dias et al. 2010)

Nevertheless, any good and even mathematically solid methodology may fail if the setup is not thoroughly planned. The performance evaluation may be hindered due to an inadequate experimental design. It is, therefore, noteworthy that although the next two chapters present multiple strategies to evaluate the performance of golf putting, the applicability of such methods depends highly on having an adequate experimental design.

Footnotes

References

  1. Battig WF (1966) Facilitation and interference. In: Bilodeau EA (ed) Acquisition of skill. Academic Press, New YorkGoogle Scholar
  2. Bjork RA (1994) Memory and metamory considerations in the training of human beings. In: Metcalfe J, Shimamura A (eds) Metacognition: knowledge about knowing. MIT Press, CambridgeGoogle Scholar
  3. Bjork RA (1999) Assessing our own competence: heuristics and illusions. In: D Gopher, Koriat A (eds) Attention and performance XVII. Cognitive regulation of performance: interaction of theory and application. MA: MIT Press, CambridgeGoogle Scholar
  4. Dias G, Mendes R, Luz M, Couceiro, MS, Figueiredo C (2010) Procedimentos e análise de dados aplicados ao estudo do putting. Boletim da SPEF 35(4):47–58. ISSN 1646-8775Google Scholar
  5. Couceiro MS, Dias G, Martins FML, Luz JM (2012) A fractional calculus approach for the evaluation of the golf lip-out. Signal Image Video 6(3):437–443. doi: 10.1007/s11760-012-0317-1 CrossRefGoogle Scholar
  6. Couceiro MS, Dias G, Mendes R, Araújo D (2013) Accuracy of pattern detection methods in the performance of golf putting. J Motor Behav 45(1):37–53. doi: 10.1080/00222895.2012.740100 CrossRefGoogle Scholar
  7. Davids K, Button C, Bennett SJ (2008) Dynamics of skill acquisition—a constraints-led approach. Human Kinetics Publishers, ChampaignGoogle Scholar
  8. Dias G, Mendes R (2010) Efeitos do contínuo de níveis de interferência contextual na aprendizagem do “putt” do golfe. Rev bras cienc esporte 24(4):545–553. doi: 10.1590/S1807-55092010000400011
  9. Dias G, Mendes R, Couceiro MS, Figueiredo C, Luz JMA (2013) “On a Ball’s trajectory model for putting’s evaluation”, computational intelligence and decision making—trends and applications, from intelligent systems, control and automation: science and engineering bookseries. Springer Verlag, LondonGoogle Scholar
  10. Dias G, Couceiro MS, Barreiros J, Clemente FM, Mendes R, Martins FM (2014) Distance and slope constraints: adaptation and variability in golf putting. Mot Control 18(3):221–243. doi: 10.1123/mc.2013-0055 CrossRefGoogle Scholar
  11. Endres F, Hess J, Engelhard N, Sturm J, Cremers D, Burgard W (2012) An evaluation of the RGB-D SLAM system. In: 2012 IEEE International Conference on robotics and automation (ICRA) pp 1691–1696Google Scholar
  12. Fernández-Baena A, Susín A, Lligadas X (2012) Biomechanical validation of upper-body and lower-body joint movements of kinect motion capture data for rehabilitation treatments. In: 4th International Conference on intelligent networking and collaborative systems (INCoS) pp 656–661Google Scholar
  13. Garafalo P (2010) Development of motion analysis protocols based on inertial sensors, Ph.D. thesis on Bioengineering, University of BolognaGoogle Scholar
  14. Guadagnoli MA, Lee TD (2004) Challenge point: a framework for conceptualizing the effects of various practice conditions in motor learning. J Motor Behav 36(2):212–224. doi: 10.3200/JMBR.36.2.212-224
  15. Harbourne RT, Stergiou N (2009) Movement variability and the use of nonlinear tools: principles to guide physical therapist practice. JNPT Am Phys Ther 89(3):267–282CrossRefGoogle Scholar
  16. Magill RA (2011) Motor learning and control: concepts and applications. McGraw-Hill, New YorkGoogle Scholar
  17. Neal RJ, Wilson BD (1985) 3D kinematics and kinetics of the golf putting. Int J Biomech 1(3):221–232. ISSN (online): 1543-2688Google Scholar
  18. Neal RJ, Lumsden R, Holland M, Mason B (2007) Body segment sequencing and timing in golf. In: Jenkings S (ed) Annual review of golf coaching 2007. Multi-Science Publication, BrentwoodGoogle Scholar
  19. Porter JM, Magill RA (2005) Practicing along the contextual interference continuum increases performance of a golf putting task. J Sport Exerc Psy 27:S-124. ISSN. 0895-2779Google Scholar
  20. Shea JB, Morgan RL (1979) Contextual interference effects on the acquisition, retention and transfer of motor skill. J Exp Psychol Learn Mem Cogn 5(2):178–187. doi: 10.1037/0278-7393.5.2.179
  21. Shotton J, Sharp T, Kipman A, Fitzgibbon A, Finocchio M, Blake A, Moore R (2013) Real-time human pose recognition in parts from single depth images. Commun ACM 56(1):116–124CrossRefGoogle Scholar
  22. Tani G (2005) Comportamento motor: aprendizagem e desenvolvimento. São Paulo, GuanabaraGoogle Scholar
  23. Ye M, Zhang Q, Wang L, Zhu J, Yang R, Gall J (2013) A survey on human motion analysis from depth data. In: time-of-flight and depth imaging. sensors, algorithms, and applications Springer, Berlin Heidelberg pp. 149–187Google Scholar

Copyright information

© The Author(s) 2015

Authors and Affiliations

  1. 1.Sport Sciences and Physical EducationUniversity of CoimbraCoimbraPortugal
  2. 2.Ingeniarius, LdaMealhadaPortugal

Personalised recommendations