Site Loader

                                                                                                          I.       
Classifier Design

    We develop a General Type-2 Fuzzy Set
(GT2FS) induced classifier to classify the n
dimensional features into three odor concentration classes. A GT2FS is a three tuple, where x is a
linguistic variable (here, feature value),  is the primary MF and is the secondary MF. Both the primary and the secondary MFs
lie in 0,1. Here, we present a GT2FS in vertical slice form 15 where each
vertical slice represents secondary MF with respect to primary membership at a
given  say.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

    Given a GT2FS classifier rule Rj
: If x1 is , x2
is , ……, xm
is  Then y is where xi is
for i=1 to mare GT2FS-induced propositions and y isdenotes a type-1 fuzzy proposition, representing a class
description. Suppose we have the measurements: x1=x1/,
x2 = x2/, …,xm= xm/.We
obtain the firing strength of rule Rj
in two steps. First, we compute the T-norm of the z-slices:  for i = 1 to m. For am-propositional IF-part of the rule j, where each z-slice contains n points, we have nm elements in the T-norm. In
the second step, we select the largest element from these nm t-norms.

 

Step-I: Compute

 

defined  by a
set {tt…t

 

Here, we take t=min.

 

Step-II: Obtain the maximum of
the elements of the set and call it firing strength (FSj) of rule Rj.

 

 (5)

Inferred
MF of the type-1fuzzy set is then obtained by

 

                             
(6)

De-fuzzification
of firing strength of rule j,

                                 (7)

We select
the rule with the highest DFj (Fig.
1).

                                                                                                
II.      
Experiments and Results

This section mainly involves in experimental framework and
three following experiments: selection of brain regions, selection of EEG
features, classifier performance.

A.   Experimental
framework

Our experiment has been done at
Artificial Intelligence Lab, Jadavpur University. We use 21-channel EEG device,
having a sampling rate of 500 Hz and resolution of 100µ in our experiment. Fig
3 depicts the entire experimental setup. Ten healthy volunteers are
participating in above experiments. Out of the 10 healthy volunteer, 4 men and
6 women in the age group 18-32 years are selected. They are requested to sit on
a comfortable chair with back- and elbow-rests in a rest condition without
movement to eliminate possible contamination by movement-related artifacts. Here,
in our experiment we use perfume, Dettol, Vinegar
(acetic acid, CH3COOH) and alcohol as stimulus with three concentration
levels; Low (25% aroma with
75% water), medium (50% aroma and 50% water) and High (75% aroma and 25%
water). All the volunteers are asked to take the odour samples with three
different concentrations (low, medium, high) respectively with a 20 seconds
time gap.To avoid the residual effect of the previous stimulus, a
time-gap of 20 seconds is maintained between two successive presentations of
odor stimuli. Each experiment with one sample presentation and EEG recordings
for one subject together constitutes one trial. Each trial is repeated 10 times
over each subject. The sequence of presentation of the odor concentration is
randomly selected in each trial from the list: low, medium and high. Fig. 4
represents the sequence of odor stimuli.

 

 

Figure 3. An Experiment set up

 

B.   Selection
of Brain Regions

EEG signals are collected from 21 electrodes, which are
placed on the scalp of the subjects. EEG signals are recorded on a separate
computer having 8 GB RAM with CPU

 

Low

Medium

High

Activation level

Odor 1 (Perfume)

 

 

 

 

Odor 2 (Dettol)

 

 

 

 

Odor 3 (Acetic acid)

 

 

 

 

Odor 4 (Alchohol)

 

Figure.5. Section of brain region and
frequency bands from scalp maps of three concentration level of four stimuli

clock of 3.4 GHz. Fig. 5 shows the scalp maps for one subject. Scalp
maps of different concentration level for different stimuli have been recorded
during the experiment. From the scalp maps, it
can be observed that pre-frontal, frontal, parietal, and temporal lobes exhibit
significant activations during the experiment. Here,
the pre-frontal region is hardly activated in olfactory sensing and processing.
Besides these, frontal and 

Post Author: admin

x

Hi!
I'm Lewis!

Would you like to get a custom essay? How about receiving a customized one?

Check it out