Information Technology Reference
In-Depth Information
In order to define the quality of a gait based on the validity of the NL propositions
calculated previously, we have designed a FRBS which captures the expert knowl-
edge thanks to its semantic expressiveness, using linguistic variables [20] and rules
[9]. Thanks this FL formalism, when designing the fuzzy rules and linguistic labels,
we can deal with isolated events of lack of symmetry or homogeneity due to turns,
steps or stumbles. However, the existence of repetitive lacks of symmetry and ho-
mogeneity represents the presence of gait disorders that must be identified by the
system. Therefore, the following set of fuzzy rules is aimed at qualifying the gait of
a person by taking into account these issues:
R
1
: If most times the homogeneity is low and most times the symmetry is low, then
the quality is low.
R
2
: If most times the homogeneity is medium and most times the symmetry is
medium, then the quality is medium.
R
3
: If most times the homogeneity is high and most times the symmetry is high,
then the quality is high.
R
4
: If few times the homogeneity is high and few times the symmetry is high, then
the quality is low.
R
5
: If few times the homogeneity is low and few times the symmetry is low, then
the quality is high.
R
6
: If some times the homogeneity is low and some times the symmetry is low, then
the quality is medium.
R
7
: If some times the homogeneity is high and some times the symmetry is high,
then the quality is medium.
The output of the FRBS is calculated as a weighted average of the individual rules.
The weight of each rule is calculated from its firing degree. To calculate this firing
degree, we use the minimum for the
and
between the two antecedents, e.g., if the NL
propositions “most times the homogeneity is low” and “most times the symmetry
is low” have a validity degrees of 0.75 and 0.5 respectively, the firing degree of
R
1
will be 0.5. Therefore, thanks to use of the FL formalism, we can express together
with the total output the reasons of why this output is obtained.
As a practical example, consider the validity degrees of the NL propositions dis-
played in Table 11.1 which were obtained during 20 complete cycles for a certain
person. With these values, only rules
R
3
,
R
5
,and
R
7
are fired with firing degrees of
Table 11.1
Example of the validity degrees of the set of NL propositions
few times some times most times
homogeneity is low
1
0
0
homogeneity is medium
0.62
0.38
0
homogeneity is high
0
0.38
0.62
symmetry is low
1
0
0
symmetry is medium
0.06
0.94
0
symmetry is high
0
0.94
0.06
