Displacement velocity [relative value] 
-0.8 + —o— /x —^— Vy 
—Dm yz —c- Vt 
  
  
  
1 3a Bin ils Ohollinddiely 170-9 
Time [frame number] 
(e) bend backward 
  
     
Displacement velocity [relative value] 
—O— Vx —N— Vy 
—L yz c Vt 
  
  
  
1 3 5 7. 9 11 13 16 8-19 
Time [frame number] 
(f) bend forward 
Figure 6 Templates of head motions. 
Let d(i, /) be the distance between vector a; and vector 
b,, then d(i, / ) is calculated by the following equation. 
d(i.j)- a, -b,| 
AW l-v lY +Vof]-v,l1} 
«ull-vvull-vauyb o 
where 
(.. [i].v,, [i] , v.i] Va [i]) 
Veli] Vi vll vl) 
Let D(T, R ) be the distance between pattern T and 
pattern R, then D(T, R ) is calculated by the following 
equation. 
a, 
b, 
. / d j, H 
D(T.R)- pe en (12) 
The template pattern which gives the least distance 
between patterns is judged as the input motion pattern. 
The recognition experiment of the head motions were 
done for the specified person whose action patterns are 
adopted as the templates and 3 unspecified persons. 
Table 2 shows the results of the matching distances 
between the unknown input patterns and templates in the 
case of the specified person. When the real actions were 
the same kind as the templates, the matching distances 
became shortest. Tabel 3-5 show the results of 
matching distance in the cases of unspecified persons. 
In these cases, the matching distances also became 
shortest between the same kind of action. 
Table 2 Distances between input action patterns of 
specified person and templates. 
  
emplate Kee 
p i bend bend 
input still nod |shake/| tilt backward| forward 
  
keep still | 0.04 | 0.25 | 0.42 |0.29| 0.34 0.26 
  
nod 0.36 | 0.22 | 0.53 |0.48 | 0.50 0.44 
  
shake | 0.35 | 0.46 | 0.25 | 0.51 0.54 0.47 
  
tilt 0.29 | 0.43 | 0.53 | 0.11 0.46 0.39 
  
  
  
  
  
  
  
  
bend 
backward 0.41 | 0.54 | 0.64 | 0.53 0.16 0.63 
bend 0.23 | 0.33 | 0.46 | 0.39 0.48 0.14 
forward 
  
Table 3 Distances between input action patterns of 
unspecified person À and templates. 
  
emplate kee 
p bend bend 
input still nod, | shake tit backward| forward 
  
keep still | 0.05 | 0.24 | 0.40 |0.26| 0.34 0.23 
  
nod 0.42 | 0.29 | 0.60 |0.53| 0.49 0.54 
  
shake | 0.36 | 0.59 | 0.26 | 0.61 0.71 0.61 
  
tilt 0.35 | 0.47 | 0.61 | 0.15] 0.50 0.47 
  
  
  
  
  
  
  
  
bend 
backward] 0:38 | 0.46 | 0.67 [0:50 | 0.17 | 0.61 
bend 10.26|0.32| 0.44 |0.39| 052 | 0.15 
forward 
  
440 
keep still | 
shake | 
backward] 
forward 
mmm 
  
Table4 I 
l 
emplate 
input 
nod 
tilt 
bend 
bend 
  
Table5 | 
| 
emplate 
input i 
keep still | 
nod 
shake 
üt | 
bend 
backward| 
bend 
forward | 
  
As a result 
recognized 
velocity ve 
available 1 
unspecifiec 
processing 
the matchi 
processing 
motions in 
We have p 
situations 
subjects in 
the perplex 
motions. 
interest arc 
motion in r 
input mot 
calculated 
it was ma 
behaviors 
available 
unspecifiex 
correctly, i 
measures 
this progre 
can be ap| 
responses 
perplex siti