Investigating Relationship between Running Motions and Skills

Acquired from Jump Trainings

Chanjin Seo

, Masato Sabanai

, Hiroyuki Ogata

and Jun Ohya

Department of Modern Mechanical Engineering, Waseda University, 3-4-1, Ookubo, Shinjuku-ku, Tokyo, Japan

Faculty of Science and Technology, Seikei University, 3-3-1, Kichijoji-kitamachi, Musashino-shi, Tokyo, Japan

Keywords: Running Motion, Jump Training, Skill, Coaching System, Stepwise Skill Improvement.

Abstract: To identify the difference in performers' motions, this paper investigates the relationship between running

motions and the result of evaluating motions during jump training. To clarify the relationship, two

experiments were performed using 17 subjects as follows: i) obtaining sequences of human joints during

running to evaluate running motions, and ii) obtaining motions during jump training which could skill up the

running motions. According to the result of those experiments, we confirmed that whether a running motion

is good or not relies greatly on the number of acquired skills.

1 INTRODUCTION

In recent years, emerging technologies such as deep

learning and image processing have made it possible

precisely to recognize objects or to detect human

poses. These technologies permit to develop com-

puterized coaching systems that obtain sports motion

data using sensors and analyze them to objectively

evaluate the learner’s performance, and to help the

learner improving skills without human coaching.

Traditional coaching system normally outputs a

one-dimensional evaluation result such as a score for

an exercise (Pirsiavash et al., 2014 and Parmar et al.,

2016) or a binary evaluation such that whether the

motion has achieved the ideal motion using sensors

(Ozaki et al., 2016). In particular, Pirsiavash et al.’s

method drew arrows on the video image to show the

direction to the ideal pose, while Ozaki et al.’s

method gave the performer a real-time voice

instruction so that the performer can improve his/her

motion. However, such systems are not always

suitable for low-level learners. One reason is that such

learners are considered not to have enough skills to

improve their performance. Another reason is that

they cannot adequately perfrom a motion along the

improvement strategy proposed by the system.

To solve such beginners' problem(s), we are

addressing to develop a coaching system that can

improve skill step by step by detecting the skills

acquired by a learner, and by automatically outputing

the improvement strategy which is appropriate for the

learner’s skill level. Our basic idea is that the system

can output a strategy to improve few problems which

cause low performance rather than to improve all the

problems. Also, we suppose that the few problems for

a learner can be solved by acquiring some skills

which he/she does not have. Therefore, we propose

two methods to resolve these problems: i) the system

finds a performer whose level is slightly higher and

has similar skill for the learner, and ii) a learner

improves his/her skill to achieve the slightly higher-

level performer’s skill.

To achieve the method i), we first focus on how to

extract and classify running skills from motions

without a priori knowledge using our previous

unsupervised learning based method (Seo et al.,

2019). However, we have not yet resolved whose

level is higher and whose skill is similar to the learner.

Note that this paper deals with training motions which

are related to running motions. The reason is that the

training motions are helpful to understand what skills

a learner has. In particular, “Skills” reflect a person’s

proficiency in performing a paricular task (Schmidt et

al., 2000). Based on the skills, we assume that

performance of a learner relies greatly on the number

of learner’s acquired skills as shown in Fig. 1, and that

a performer, whose level is slightly higher and who

has skills similar to the learner, has more skills than

the learner (Fig. 1). In fact, we suppose the learner in

Beginner level doesn’t have some skills even to

perfrom basic trainings related to running motions.

198

Seo, C., Sabanai, M., Ogata, H. and Ohya, J.

Investigating Relationship between Running Motions and Skills Acquired from Jump Trainings.

DOI: 10.5220/0008348301980203

In Proceedings of the 7th International Conference on Sport Sciences Research and Technology Support (icSPORTS 2019), pages 198-203

ISBN: 978-989-758-383-4

Table 1: Observational motion evaluation items concerning running motion in sprinting: These evaluation items can be

evaluated for performers by keeping scores from Suzuki et al.’s items.

Score

1.0

0.5

0.0

Size of lower

limb move-

ment

The knee of the

swinging leg moves

forward largely, and the

leg swings back just

below the body.

The swinging legs are swinging weakly in

running motion.

The forward swinging of the

swinging leg and the extention

of the knee are very small, and

the flight duration is extremely

short.

There is no swing back of the swinging leg

in the direction directly beneath the body,

and the foot of the swinging leg is

touching the ground immediately before

going in front of the swinging legs.

Switching of

legs

The swinging leg

overtakes the supporting

leg almost at the same.

The swinging leg overtakes the supporting

foot immediately after touching the ground

The swing leg slowly

overtakes the supporting leg

by touching the ground

And also, our opinion is that the learner should

acquire some skills from the basic trainings in order

to perform a running motion such as the performer.

To confirm that the performance is affected by the

number of acquired skills, this paper investigates the

relationship between running motions and motions

during training, which could skill up running motions.

Figure 1: Image of skills acquired by each performer. The

number of skills acquired by a lower-level performer is

generally smaller than an expert.

2 METHODS

To confirm what running skills each performer has,

we conduct the following two experiments: i) to

obtain sequences of human joints during running

motions and to evaluate the running motions (Section

2.1), and ii) to obtain motions during jump trainings,

which could improve the running motions (Section

2.2).

For these experiments, we collected data from 17

healthy male subjects (10 beginners and 7 experts;

21-24 years old).

2.1 Running Motions

To collect data, we let each subject run on a 30 m

track at a full speed. A video camera was placed at the

position shown in Fig. 2 to capture the first 15 m

running motion in order to check the start dash. The

videos were taken at 60 frames per second. Each

subject was asked to run 4-9 times, and 98 videos

were recorded in total.

Figure 2: Left: Environmental setup for getting running

motion data. Right: An example of the images captured by

the video camera.

The running motions were evaluated using Suzuki

et al.’s items by an expert sprinter. We especially

focus on the two items related to the low limb, which

are shown in Table 1. The period of evaluating those

items begins when the supporting leg leaves the

ground and ends when the opposite foot reaches the

ground. The score is obtained as an average rating of

each period in the running motion.

2.2 Jump Training Motions

To estimate each subject’s skill(s), we collected the

data of trainings, which could skill up the running

motion. The experimental condition is shown in Fig.

Training items were selected from Tanaka’s book

(Takano, 2008) on training for sprinters. The selected

items are shown in Table 2. The reason for selecting

these items is that jump trainings such as skip is

related to skill up running motions according to the

report of Kotzamanidis and Iwatake et al., and it is not

difficult even for beginners to simply perform such

jump trainings.

Investigating Relationship between Running Motions and Skills Acquired from Jump Trainings

199

The subjects were thoroughly explained about

the methods of performing the jump trainings before

conducting the experiment. Then, a two-stage

evaluation of whether the trainings of the items in

Table 2 were performed well or not was judged by the

same expert sprinter as Section 2.1.

Figure 3: Left: Environmental setup for getting training

motion data. The stick ladders were placed at the interval of

0.4 m and the subjects jumped the interval in performing

the training items in Table 2. Right: An example of the

images captured by the video camera.

Table 2: Training Items in performing the experiment. In

particular, in performing jump trainings labeled 2, 7 and 8,

our subjects skip backward one step after two steps forward

in Fig. 3’s experimental condition.

Label

Training Name

Skip forward using both legs

Skip forward and backward using both legs

Skip forward using left leg

Skip forward using right leg

Sideways skip using left leg

Sideways skip using right leg

Skip forward and backward using left leg

Skip forward and backward using right leg

3 RESULTS

3.1 Evaluating Running Motions

The scores obtained in the first experiment is shown

in Table 3. From the table, it can be seen that almost

all the score of the beginners are below 0.5 in Items 0

and 1. We could not confirm other common tendency

of the beginners except that they were evaluated low

in Table 1.

On the other hand, it can be seen that almost all

the score of the experts are over 0.5 for Items 0 and 1,

and also the standard deviations of the experts are

smaller than those of the beginners in Table 3.

Expert0 got a score lower than other experts. That is

because Expert0 had not run for a long time, which

results in losing his past capability.

As a result, we could confirm that the running

motions are different between beginners and experts

through Suzuki et al.’s evaluation method. However,

the reason why their motions are different cannot be

clarified only by scores in Table 3.

Table 3: Scores of each subject: the number of times to run,

mean score, and standard deviation (SD) for Table 1’s

items.

Subject

Times

Item0

Item1

Mean

Beginner0

0.27

0.11

0.39

0.04

Beginner1

0.55

0.18

0.31

0.12

Beginner2

0.07

0.08

0.03

0.02

Beginner3

0.22

0.10

0.28

0.13

Beginner4

0.38

0.20

0.27

0.19

Beginner5

0.30

0.10

0.23

0.08

Beginner6

0.25

0.05

0.42

0.14

Beginner7

0.31

0.16

0.32

0.10

Beginner8

0.16

0.08

0.02

0.04

Beginner9

0.36

0.08

0.27

0.06

Expert0

0.66

0.11

0.53

0.17

Expert1

0.98

0.02

0.91

0.03

Expert2

0.93

0.04

0.94

0.06

Expert3

0.98

0.04

0.94

0.03

Expert4

0.88

0.08

0.87

0.03

Expert5

1.00

0.00

0.77

0.00

Expert6

0.94

0.02

0.81

0.08

Average

5.76

0.55

0.33

0.49

0.31

3.2 Evaluating Jump Training Motions

Figure 4 and Table 4 show the relationship between

scores of Item 0 in Table 3 and the result of evaluating

jump training motions. The subjects are divided into

four groups based on the scores of Item0. From Fig.

4 and Table 4, it appears that subjects who are in a

similar skill level tend to belong in the same group.

For example, the result of evaluating jump training

motions is low if the Item0’s score is low, and the

result is high if it is high. In particular, in Fig. 4, the

subjects of Area1 in Table 4 tend to be highly possible

to perform Item7 in Table 2 rather than others, and

Area2 tends to be highly possible to perform certain

items in Table 2 but difficult to perform others.

Figure 5 and Table 5 show the relationship

between scores of Item 1 in Table 3 and the result of

evaluating jump training motions. In Fig. 5 and Table

5, though the scores are divided at regular interval

into 5 groups, we merge the ranges of 0.6 ≤ 𝑥 < 0.8

and 0.8 ≤ 𝑥 < 1.0 because only experts belong to

these ranges. Therefore, there are 4 groups as a total.

Each area in Table 5 tends to contain subjects who

have similar skill levels. In particular, Area2 in Table

2 is composed of a beginner and an expert in Fig. 5.

Thus, the result in Fig. 5 depends on the expert’s

result and leads to show a better result than Area3 in

icSPORTS 2019 - 7th International Conference on Sport Sciences Research and Technology Support

200

Item6 and Item7, but, in Table 5, it is confirmed that

mean score is different.

Figure 4: The evaluation result of whether each training

items in Table 2 can be performed or not by subjects in

anarea which is decided by a mean score in Item0 of Table

3. In the figure, x is the mean score in Item0 of Table 3.

Table 4: The basic information in each area: the number of

persons that belong to each area, mean score, and standard

deviation (SD) of evaluated result in all training items. The

range of Areas 0 to 3 are 0 ≤ 𝑥 < 0.25, 0.25 ≤ 𝑥 < 0.5,

0.5 ≤ 𝑥 < 0.75, and 0.75 ≤ 𝑥 , respectively. 𝑥 is a mean

score of Item 0 in Table 3.

Area0

Area1

Area2

Area3

Belonged persons

Mean

0.03

0.40

0.69

0.90

0.08

0.26

0.24

0.08

Figure 6 and Table 6 show the relationship

between total scores of Items 0 and 1 in Table 3 and

the result of evaluating jump training motions. In Fig.

6 and Table 6, the scores are divided into 4 groups.

Each area in Table 6 tends to contain subjects who

have similar skill levels. In particular, Area2 and

Area3 in Table 3 are composed of all the experts, and

it can be confirmed that some items in Table 3 could

not be performed even experts in Area2 from Fig. 6.

Figure 5: The evaluation result of whether each training

items in Table 2 can be performed or not by subjects in an

area which is decided by a mean score in Item1 of Table 3.

In this figure, x is the mean score in Item1 of Table 3.

Table 5: The basic information in each area: the number of

persons that belong to each area, mean score, and standard

deviation (SD) of evaluated result in all training items. The

range of Areas 0 to 3 are 0 ≤ 𝑥 < 0.2 , 0.2 ≤ 𝑥 < 0.4,

0.4 ≤ 𝑥 < 0.6 , and 0.6 ≤ 𝑥 , respectively. 𝑥 is a mean

score of Item 1 in Table 3.

Area0

Area1

Area2

Area3

Belonged persons

Mean

0.00

0.32

0.63

0.90

0.00

0.17

0.22

0.08

Figure 6: The evaluation result of whether each training

items in Table 2 can be performed or not by subjects in an

area which is decided by a total score in the mean score of

Item0 and Item1 in Table 3. In this figure, x is the total score

in the mean score of Item0 and Item1 in Table 3.

Table 6: The basic information in each area: the number of

persons that belong to each area, mean score, and standard

deviation (SD) of evaluated result in all training items. The

range of Areas 0 to 3 are 0 ≤ 𝑥 < 0.6, 0.6 ≤ 𝑥 < 1.2 ,

1.2 ≤ 𝑥 < 1.8, and 1.8 ≤ 𝑥 ≤ 2, respectively. 𝑥 is a total

score of Item0 and Item1 in Table 3.

Area0

Area1

Area2

Area3

Belonged persons

Mean

0.03

0.48

0.79

1.00

0.08

0.19

0.16

0.00

4 DISCUSSION

In Section 3.2, it was observed that there is a

relationship between the skill to perform certain jump

training and the scores in Table 3. In particular, the

total scores of Item0 and Item1 in Table 3 show the

difference in each area of whether each training item

can be performed or not from Fig. 6, and the areas of

Table 6 show a stepwise distribution between the

scores and the result of training motions. From these

results, we can infer that the running motion is related

by the number of obtained skills from the jump

trainings in Table 2. Then, the number of acquired

skills highly affects the motions of the runner.

-0.2

0.2

0.4

0.6

0.8

1.2

1 2 3 4 5 6 7 8

Score

Each Training Item

0<=x<0.6 0.6<=x <1.2 1.2<=x < 1.8 1.8<=x <=2

Investigating Relationship between Running Motions and Skills Acquired from Jump Trainings

201

However, all the motions are evaluated from their

appearances in this paper. Thus, we could obtain the

relationship about skills, but we could not validate the

motions using detail information such as the motion

sequence of human joints.

In our future works, we will focus on validating the

training motions by detecting human joints, and we

will clarify the motion difference which causes the

different numbers of obtained skills.

5 CONCLUSIONS

In this paper, we have investigated the relationship

between the running motions and the motions during

jump trainings so as to clarify the difference in running

motions of performers which causes different numbers

of acquired skills. From our experiments, we can infer

that the number of acquired skills from jump trainings

is related to the performance of running motions.

ACKNOWLEDGEMENTS

This study was conducted as a part of research

activities of the Human Performance Laboratory,

Organization for University Research Initiatives,

Waseda University.

REFERENCES

Iwatake, J., Kitada, K., Kawahara, S., Zushi, K., 2008. The

influence of jump training on the sprinting ability in late

pubertal boy students. Japan Journal of Physical

Education, Health and Sport Sciences, 53, 352-362.

Kotzamanidis, C., 2006. Effect of plyometric training on

running performance and vertical jumping in prepubertal

boys, J Strength Cond Res, 20(2), 441-445.

Ozaki, A., Honda, M., 2016. Development of an Automated

Motion Coaching System Using a Portable Sensor, Japan

Society for Information and Systems in Education, 33(1),

22-30.

Parmar, P., Morris, T., B., 2016. Measuring the quality of

exercises. In IEEE International Conference Engineer-

ing in Medicine and Biology Society, 556-571.

Pirsiavash, H., Vondrick, C., Torralba., A., 2014. Assessing

the Quality of Actions. European Conference on

Computer Vision (ECCV), 556-571.

Schmidt, A., R., Wrisberg, A., C., 2000. Motor Learning and

Performance, Human Kinetics. USA, 2

edition, 29-31.

Seo, C., Sabanai, M., Ogata, H., Ohya, J., 2019.

Understanding Sprinting Motion Skills Using

Unsupervised Learning for stepwise Skill Improvements

of Running Motion. In 8th International Conference on

Pattern Recognition Application and Methods

(ICPRAM), 467-475.

Suzuki, Y., Tomozoe, H., Yoshinaga, T., Kaji, M., Hirayam,

K., 2016. Study on the observed behaviour evaluation

method of sprinting action (in Japanese). Japan Journal

of Physical Education, Health and Sport Sciences, 32(1),

1-20.

Takano, S., 2008. Track and Field, Sprinting Perfect Master,

Sports Step-Up DVD Series (in Japanese), Shinsei

Publishing Company, Japan, 33-35.

APPENDIX

We show the images of trainings in Table 2. In first,

label 1-4, 7 and 8 in Table 2 were performed such as

Fig. 7. In the case of the training of label 1, 3 and 4,

subjects jump over a stick ladder until the end.

Figure 7: Image of skipping forward and backward.

In case of the training of label 2, 7 and 8, the

subjects performed 2 steps. First step is jumped over a

stick ladder twice in Fig.7 (1 and 2). Next step is

jumped over a stick ladder backward in Fig.7 (3). And,

the subjects repeat to do this motion until the end.

Figure 8: Image of Sideways skip.

icSPORTS 2019 - 7th International Conference on Sport Sciences Research and Technology Support

202

On the other hands, label 5 and 6 in Table 2 were

performed such as Fig. 8. In the case of the training

of label 5 and 6, the subjects jump over a stick ladder

until the end.

Investigating Relationship between Running Motions and Skills Acquired from Jump Trainings

203