Page 134 - Control Theory in Biomedical Engineering

Genetic fuzzy logic based system for arrhythmia classification 119

Table 5 Rule base.
Rule AND AND AND AND AND Then
no If E1 is E2 is E3 is E4 is E5 is E6 is Sis

R1 MOY MOY MOY MOY MOY MOY NSR
R2 MIN MIN – MAX – – PVC
R3 MOY – – MAX MAX – P
R4 MIN MAX MOY MAX – MAX RBBB
R5 MIN MIN MOY MAX – – LBBB

2.2.2 FLC optimization
Unlike standard controllers, the FLC configuration requires the adjustment
of a greater number of parameters, as discussed in the previous section. In
fact, the FLC designer has to make tuning regarding the expression of the
rules, the definition of inputs and its fuzzy values, the inference mechanism,
the defuzzification method and many others. Thus, configuring the FLC
with the appropriate parameters is a challenging task, especially when expert
knowledge is not available. Hence, optimizing the FLC parameters offers a
reliable solution to this problem. We have applied the GA for the optimi-
zation of the Gaussian membership parameters and the rules number.
The GA is applied for optimization tasks and it is usually used in problems
with amount parameters. By using genetic optimization, it is essential to
define the chromosome representing the solution and the fitness function
evaluating the produced solutions.
As it is described in Fig. 12, the GA process begins by using the initial set
of solutions, named initial population, which is randomly generated and is
subsequently coded into binary chromosomes. Then, the FLC is updated
with each chromosome and consequently evaluated using a suitable fitness
function. In this study, we have selected the Root Mean Square Error
(RMSE) as a fitness function (see Eq. (9)).

v ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u N
u 1 X 2
RMSE ¼ t ð t i o i Þ (9)
N
i¼0
Thus, the best solution is a vector that reaches the minimal value of the
RMSE function. Indeed, the predicted output (o i ) is compared with its cor-
responding target (t i ), in order to evaluate the RMSE function. After that, in
order to create the diversity of the population, selection, crossover and
mutation operators are applied and a new population is evaluated
(Lassoued and Ketata, 2018a). This process continues for multiple iterations