Digital Signal Processing Reference
In-Depth Information
(a )ViSOR dataset
(b)
Weizmann dataset
Fig. 2.
Exam
mple of motion descriptors for human activities
2.5
SVM Classification
Classification was performe
sifier has been successfully
bustness, applicable results
one-against-one SVM mu
classes,
ed using a bank of binary SVM classifiers. The SVM c
y used in many pattern recognition problems given its
s and efficient time machine. In our approach, we use
ulticlass classification
[12], where given
mot
are built and the best class is selected by a voting strate
ned with a set of motion descriptors, extracted from h
equences (see next section). The Radial Basis Funct
].
las-
ro-
the
tion
egy.
hand
tion
classifiers a
The SVM model was train
labeled human activity se
(RBF) kernel was used [13]
