Soft Entrainment: Co-emergence of “Maai” and Entrainment by Rhythm Controller - Emotional Engineering - page 81

Consequently, for the avatar operation with a rhythm controller for which the inputs
are Δ T and Δ A, there are many controller inputs for realizing a speci ! c avatar
motion. Avatar velocity (V) can be expressed in the following form:

V ¼ f D T ; D A ð Þ ð 1 Þ

This function f expresses the relationship between controller input and avatar
motion—in other words, the controller manipulation method. As mentioned before,
when the avatar is operated using the rhythm controller, the operator changes the
method for creating the area of the controller waveform by adjusting its cycle and
amplitude. Consequently, the operator can temporally (dynamically) change not
only avatar velocity but also the controller manipulation method (function f). In this
study, at the start of examining such a controller manipulation method, ( Δ T, Δ A, V)
was plotted in a 3D scatter plot when the zero-cross was created on the controller
waveform. Furthermore, we tried to estimate the function f by multiple regression
analysis in which the explanatory variables are Δ T and Δ A, and the dependent
variable is V.

3.3 Ambiguity of the Controller Manipulation Method

We observed how the controller manipulation method changed when the operator
performed similar avatar motions in different situations. Speci ! cally, we conducted
three different experiments, and listed here:

• Forward and backward movement experiment (Fig. 5 a )
• Chase up experiment (Fig. 5 b)
• Kendo experiment (Fig. 7 a).
The chase up experiment required the creation of action that is complementary with
the motion of the automatic avatar. However, there was no need for it in the forward
and backward movement experiment. The Kendo experiment was similar to the
chase up experiment in that action must be complementary with the motion of the
opponent avatar. But for the Kendo experiment, the plot of the match is not decided
ahead of time. Before the match, players cannot decide in what situations they will
create and collapse Maai. Players needed to improvise creating the action with the
opponent depending on the situation.
Table 1 shows analysis results of the relationship between the controller input
( Δ T, Δ A) and avatar motion (V) in the three experiments described above. From this
table, we found that the avatar velocity (V) in all experiments was approximated by
the regression plane determined from Δ T and Δ A, because the determination
coef ! cient of the multiple regression analysis was near 1 in all experiments. Table 1
shows that the partial regression coef ! cient of the forward and backward movement
experiment, in which subjects do not need to coordinate avatar motion with the
movement of an opponent, differs from the chase up and Kendo experiments, in