**6.1 Comparison of time and error**

The experimental data of four subjects could not be used due to the subjects' misunderstanding of tasks and some mistakes in the system setting. The data of another three subjects were statistically eliminated. Of these three, two subjects moved the grip too quickly and had large errors, and another subject moved the grip too slowly and took a long time to play the game. Thus, the data of a total of 20 subjects were compared for duration time, grip (cursor) position error from the center of the wave line, and subject age.

Fig. 11 shows the relationship between the duration time and age with the three variations of applied haptic force. The horizontal axis is expressed as the log of the age. No significant correlations are confirmed. For example, the highest correlation, R2 = 0.0029, is in the case of "no force applied".

Haptic Device System for Upper Limb and Cognitive

R² = 0.400

No Force Appliced

Fig. 13. Relationship of duration time and RMS error of the grip position

increased decreased

results, and so the training effect can be excluded from consideration.

Time

increased 1 11

decreased 6 2 RMS error

and (b): "No Force Applied" and "Assistance".

Viscosity Assistance

0

5

10

15

20

Error R.M.S (mm)

25

30

35

Rehabilitation – Application for Development Disorder Children 161

regression line of "viscosity" is steeper than those of the other two cases. This result means that the viscosity force assists the precise grip position control. In contrast, the assistance force assists the grip movement toward the designated position, which allowed the user to move the

0 0.5 1 1.5 2

RMS error

Log Time(sec)

(a) Viscosity (b) Assistance Table 1. Number of subjects showing a difference in duration time and RMS error of the grip motion for "no force applied" and one other force. (a): "No Force Applied" and "Viscosity",

Table 1 shows the number of subjects showing a difference in duration time and RMS error for "no force applied" and another force. The experiments were always done in the same order. Thus, if the training affected the results, the subjects would move the grip more quickly (duration time decreases) and more precisely (RMS error decreases) for "no force applied". Table 1 (a) indicates the case of "viscosity". The RMS error increased and the duration time decreased for 11 subjects (55%). In the case of "assistance", shown in Table 1 (b), the RMS error increased and the duration time decreased for 15 subjects (75%). Conversely, in (a) the duration time increased for 7 subjects (35%) in the case of "viscosity" and in (b) for 2 subjects (10%) in "assistance". The RMS error decreased for none of the subjects. In other words, the relationship between the duration time and RMS error has the same tendency of the previous

The difference of duration time and the difference of RMS error of the grip position are plotted in Fig. 14. This figure confirms the high correlation of "assistance". The average

R² = 0.239

increased decreased

Time

R² = 0.464

increased 0 15 decreased 2 3

grip quickly. However, a short duration time generally incurs a large position error.

Fig. 11. Relationship of age of subjects and duration time

Fig. 12 shows the relationship between the root mean square (RMS) error of the grip position and the age with the same three variations of applied haptic force. The horizontal axis is again expressed as the log of the age. The correlations are negative in all cases. The RMS errors of the older subjects are lower than those of the younger subjects. The highest correlation, R2 = 0.330, is confirmed in the case of "no force applied". In the case of "viscosity", the correlation is R2 = 0.256. The case of "assistance force" has the lowest correlation, R2 = 0.102.

Fig. 12. Relationship of age of subjects and RMS error of the grip position

Fig. 13 shows the relaxation between the duration time and RMS error of the grip position with the three variations of applied haptic force. The horizontal axis is represented as the log of the duration time. All correlations are negative. This tendency is the same as that confirmed in a previous experiment by adult subjects. The correlation for "viscosity" is R2 = 0.464, that for "assistance force" is R2 = 0.400, and that for "no force applied" is R2 = 0.239. The slope of the

No Force Applied Viscosity Assistance

1.7 1.75 1.8 1.85 1.9

Log Age(Month)

R² = 0.330 R² = 0.256

Fig. 12 shows the relationship between the root mean square (RMS) error of the grip position and the age with the same three variations of applied haptic force. The horizontal axis is again expressed as the log of the age. The correlations are negative in all cases. The RMS errors of the older subjects are lower than those of the younger subjects. The highest correlation, R2 = 0.330, is confirmed in the case of "no force applied". In the case of "viscosity", the correlation is R2 = 0.256. The case of "assistance force" has the lowest

1.7 1.75 1.8 1.85 1.9

Log Age(Month)

Fig. 13 shows the relaxation between the duration time and RMS error of the grip position with the three variations of applied haptic force. The horizontal axis is represented as the log of the duration time. All correlations are negative. This tendency is the same as that confirmed in a previous experiment by adult subjects. The correlation for "viscosity" is R2 = 0.464, that for "assistance force" is R2 = 0.400, and that for "no force applied" is R2 = 0.239. The slope of the

0

Fig. 11. Relationship of age of subjects and duration time

R² = 0.102

Fig. 12. Relationship of age of subjects and RMS error of the grip position

No Force Applied Viscosity Assistance

20

40

Time (sec)

correlation, R2 = 0.102.

Error R.M.S (mm)

0

5

10

15

20

25

30 35

60

regression line of "viscosity" is steeper than those of the other two cases. This result means that the viscosity force assists the precise grip position control. In contrast, the assistance force assists the grip movement toward the designated position, which allowed the user to move the grip quickly. However, a short duration time generally incurs a large position error.

Fig. 13. Relationship of duration time and RMS error of the grip position


(a) Viscosity (b) Assistance

Table 1. Number of subjects showing a difference in duration time and RMS error of the grip motion for "no force applied" and one other force. (a): "No Force Applied" and "Viscosity", and (b): "No Force Applied" and "Assistance".

Table 1 shows the number of subjects showing a difference in duration time and RMS error for "no force applied" and another force. The experiments were always done in the same order. Thus, if the training affected the results, the subjects would move the grip more quickly (duration time decreases) and more precisely (RMS error decreases) for "no force applied". Table 1 (a) indicates the case of "viscosity". The RMS error increased and the duration time decreased for 11 subjects (55%). In the case of "assistance", shown in Table 1 (b), the RMS error increased and the duration time decreased for 15 subjects (75%). Conversely, in (a) the duration time increased for 7 subjects (35%) in the case of "viscosity" and in (b) for 2 subjects (10%) in "assistance". The RMS error decreased for none of the subjects. In other words, the relationship between the duration time and RMS error has the same tendency of the previous results, and so the training effect can be excluded from consideration.

The difference of duration time and the difference of RMS error of the grip position are plotted in Fig. 14. This figure confirms the high correlation of "assistance". The average

Haptic Device System for Upper Limb and Cognitive

they do not operate the grip and program carefully.

Rehabilitation – Application for Development Disorder Children 163

Games" group may be more familiar with computers and playing TV games. Therefore,

Subject A 10.93 30.83 4.35 32.04 2.79 30.27 Subject B 38.11 9.7 25.39 8.76 9.4 26.77 Average 26.01 11.71 23.7 12.08 16.25 16.6

Table 2. Comparison of results between subjects with negative impressions and the average

Good at Sports 23.47 14.11 23.37 13.63 11.8 18.51 Not Good at Sports 30.28 9.82 14.34 12.45 19.28 16.86 Good at TV Game 24.08 12.15 22.18 12.36 15 16.43 Not Good at TV Game 29 11.03 25.97 11.66 18.11 16.84 Average 26.01 11.71 23.7 12.08 16.25 16.6

Table 3. Comparison of results between "Good at Sports", "Not Good at Sports", "Good at

No correlation was found between the age and the duration time. The assumed reason is the range and the variation of the subjects' ages. Some younger subjects could not clearly understand the instruction, "Keep cursor (starfish) inside the thick wavy line" and operated it impulsively. Some older subjects understood the purpose of the program, "Try to keep moving the cursor on the center of the thick wavy line" and tried to do it carefully. Therefore, when the subjects are younger or they are not developing as expected, the experimenter should explain the task using easy words and be careful to make sure the subjects understand. In the comparison between "no force applied" and "viscosity" or "assistance" applied, the duration time with the "assistance" force was shorter than that of the other cases. In the "assistance" case, it was observed that the upper limb of the subjects seemed to be pulled by the haptic device. Thus, the "assistance" force may prevent voluntary motion. However, for children who have problems with voluntary upper-limb movement or hand-eye coordination, the guiding force can help move her or his upper limb (grip) to the designated point. This assistance can stimulate motion and cognitive function. The suitable level of assistance and how it is involved in therapeutic training should be discussed in the future. The error of motion decreased according to age. It was confirmed that error and age have a negative correlation. From observations, it was also confirmed that most of the older subjects tried to move the cursor along the center of the thick wavy line in the "Starfish" program. Thus, the results were reasonable. However, children at 5 years of age start to write Japanese characters and Arabic numbers as preparation for elementary school.

TV Games", and "Not Good at TV Games" groups and the average

Therefore, a simple line-drawing task should be easy for them to understand.

The negative correlation for the duration time and the grip position error was confirmed in the two different cases of applied force and no force applied. The tendency was the same as that in previous research for young adult to elderly subjects. To move the cursor in a smooth and fast manner, the subject must see the cursor, recognize its position, and move the

**7. Discussion** 

duration time RMS error duration time RMS error duration time RMS error

duration time RMS error duration time RMS error duration time RMS error

No Force Applied Viscosity Assistance

No Force Applied Viscosity Assistance

duration time of "viscosity" is -2.31 sec and that of "assistance" is -9.76 sec. The average RMS error of "viscosity" is 0.38 mm and that of "assistance" is 4.89 mm. In both cases, the significant difference was confirmed by a T-test (p<0.05).

Difference of Time (sec)

Fig. 14. Relationship between the difference of duration time and difference of RMS error of "viscosity" and "assistance"

#### **6.2 Interview**

A post-experiment interview was carried out for all subjects. 19 subjects answered the question "How was this game?" The subjects could answer freely about her or his impression of the experimental program. The vocabulary of their explanations was limited because they were young. The most common answers were "fun" (12 subjects), "interesting" (9 subjects), and "easy" (3 subjects). All subjects who gave an answer had a positive impression.

Two subjects gave both positive and negative impressions. Subject A (age 4 years 9 months, male) said, "Fun, but one of the pre-experiment programs was not interesting. It was too easy." Another subject B (age 5 years 8 months, male) said, "Fun, but I was tired." Subject A moved the grip quickly and finished tasks more quickly than average. In contrast, subject B moved slower than average and had less error than average. However, in the case of "assistance", subject B moved quickly and the error increased. This could be a sign of fatigue (Table 2).

Simple questionnaires were also used. We asked two questions: "Are you good at sports?" and "Are you good at TV games?" The subjects were categorized into a "good at" group and a "not good at" group. Table 3 shows the comparison of the group results and the average results. There are no significant differences. However, the "Good at Sports" and the "Good at TV Games" groups tend to have a shorter duration time than that of the "Not Good at Sports" or "Not Good at TV Games" groups. Due to the shorter duration time, the RMS error of the "Good at" group is larger than that of the "Not Good at" group. The "Good at Sports" group and the "Not Good at Sports" group can generally be expressed as "active" or "non-active" personality groups. The tendency of their motions may be quick, however the results are a shorter duration time and a larger RMS error. The "Good at TV


Games" group may be more familiar with computers and playing TV games. Therefore, they do not operate the grip and program carefully.

Table 2. Comparison of results between subjects with negative impressions and the average


Table 3. Comparison of results between "Good at Sports", "Not Good at Sports", "Good at TV Games", and "Not Good at TV Games" groups and the average
