A Globally Generalized Emotion Recognition System Involving Different Physiological Signals

. 2018 Jun 11;18(6):1905. doi: 10.3390/s18061905

Algorithm 2 The learning algorithm of the CNN.

1:
Input: Features of physiological signals
2:
Output: CNN Templets and $A_{g l o b a l}$
3:
Generate CNNs Templates
4:
Set $A_{g l o b a l}$ to 0
5:
for $n = 50, 100, 150, 1000, 1500$ do
6:
for $β = 0.1, 0.01, 0.001, 0.0001, 0.00001$ do
7:
for $i = 1, \dots, 10$ do
8:
Using the $i^{t h}$ learning dataset
9:
Run $S i m u l i n k M o d e l_{B}$
10:
Read Y from $S i m u l i n k M o d e l_{B}$
11:
Using Y and g
12:
Estimate $A_{g l o b a l}$ using Equation (10)
13:
Using the $i^{t h}$ learning dataset with the estimated $A_{g l o b a l}$
14:
Run SimulinkModel B
15:
Read the emotion states from $S i m u l i n k M o d e l_{B}$
16:
Calculate the Accuracy
17:
Set the Accuracy to Accuracy repository(i)
18:
end for Set the Global Parameter( $n, β$ ) = the average of the Accuracy repository
19:
end for
20:
end for
21:
Select the best CNN configuration from the Global Parameter