Soft Entrainment: Co-emergence of “Maai” and Entrainment by Rhythm Controller - Emotional Engineering - page 82

Table 1 Results of the multiple regression analysis [ 11 ]
Experiment condition Partial

regression
coef ! cient

Standardized
partial
regression
coef ! cient

Determination
coef ! cient R 2

Δ T Δ A Δ T Δ A

Forward and backward movement
experiment (Player Y)

73 71 0.5 0.54 0.97

Chase up experiment (Player Y) 18 71 0.24 0.72 0.85

Kendo experiment Player Y: winner 17 94 0.10 0.90 0.95

Player K: loser 12 95 0.23 0.76 0.88

which subjects need to coordinate avatar motion with that of an opponent. These
experimental results show that the controller manipulation method (relationship
between the controller input and avatar motion) varies with the situation.
In the forward and backward movement experiment, in which the subject does
not need to coordinate avatar motion with an opponent, data plotted in a 3D scatter
plot (Fig. 13 ) is distributed in a line. In the chase up and Kendo experiments, it is
distributed in a plane. In fact, in the forward and backward movement experiment,
the correlation coef ! cient between explanatory variables ( Δ T and Δ A) is 0.81—
nearly 1. This indicates a high likelihood of reducing the number of explanatory
variables in the multiple regression analysis. Therefore, we tried to approximate the
result of the forward and backward movement experiment with a straight line using
the analysis of the principal component as shown in Fig. 14 . From the result, the
determination coef ! cient of the regression line is 0.98—nearly 1. This result shows
the validity of approximating the relationship between the controller input and
avatar motion by simple linear regression analysis. Controller operation using the
rhythm controller has two-degree-of-freedom ( Δ T and Δ A). However, the operator
creates avatar motion by virtually reducing the number of degrees of freedom in the
forward and backward movement experiment, where the operator does not need to
coordinate avatar motion with an opponent.
As shown in Fig. 15 , in the Kendo experiment, data plotted in a 3D scatter plot is
approximated by two regression lines in the regression plane given above. Fig-
ure 15 shows the result for player Y, who won the Kendo match. Line B in Fig. 15
is the controller manipulation method in which the operator fl uctuates the Δ A, with
the Δ T kept near to constant. Figure 16 a shows a time-series variation of two
player's avatar position and the distance between two avatars. In Fig. 16 a, the zones
in which player Y created an avatar motion using line B are painted with a green
band (zones 1y - 5y). From this ! figure, we found that zones 1y - 4y correspond to just
before increasing in the distance between the two avatars. In other words, player Y
creates avatar motion using line B just before the collapse of Maai. Zone 5y
corresponds to just before decrease in the distance between the two avatars.
However, player Y swings a sword soon after Zone 5y. Therefore, we ! nd that
Zone 5y also corresponds to just before the collapse of Maai. These results show