1-1(15) 1-2(14) 
lass window, 
manually. The 
respectively, 
implementa- 
ng the recog- 
subspace is 
before. To 
mental LDA, 
incremental 
on this up- 
andle a long 
lass models. 
detected and 
one example 
r ave the sys- 
this exam- 
ocedure and 
The second 
user for the 
chy consists 
-1 and 1-2. 
two classes, 
be subspace 
ire indented 
sample pro- 
ries can not 
sked for the 
mg new in 
die learned 
Figure 3: Initialisation: (a) Initial example, (b) Result of detecting new instances, (c) Prototypes. (14): Number of associated samples, 
(d) Samples projected into the LDA subspace. Red: class 1 -1, green: class 1 -2, blue: background. Dashed lines: decision boundaries 
according the class boundaries and the rejection area.