Biomedical Engineering Reference
In-Depth Information
where
E
i
is the excitation of the
i
th neuron of the
R
-layer;
a
kj
is the excitation of the
j
th neuron of the
A
-layer in the
k
th distortion; and
w
ji
is the weight of the connection
between the
j
th neuron of the
A
-layer and the
i
th neuron of the
R
-layer. After that,
the neuron-winner is selected as a result of recognition.
Rule 2. The second rule consists of calculations of
R
-layer neuron excitations
and selection of the neuron-winner and its nearest competitor for each distortion.
For the
k
th distortion, the relation
r
k
of the neuron-winner excitation
E
wk
to its
nearest competitor excitation
E
ck
is calculated:
E
wk
E
ck
:
r
k
ΒΌ
(3.15)
After that, we select distortion with the maximal
r
k.
The neuron-winner of this
distortion is considered to be the result of recognition.
3.3 LIRA-Grayscale Neural Classifier
To adapt the LIRA classifier for grayscale image recognition we have added
additional neuron layer between the
S
-layer and the
A
-layer. We term it the
I
-layer (intermediate layer, see Fig.
3.9
).
Each input of the
I
-layer neuron has one connection with the
S
-layer. Each
output of this neuron is connected with the input of one neuron of the
A
-layer.
S-LAYER
I-LAYER
A-LAYER
R-LAYER
ON
OFF
OFF
OFF
ON
OFF
ON
OFF
Fig. 3.9. LIRA-grayscale scheme
