Page 135 - 2024-Vol20-Issue2
P. 135
131 | Fathi & Aziz
Fig. 3. Examples of the proposed dataset, eight hand images
from the same person
Fig. 4. Image renaming method
based on slops. Slop calculation formula equation 1 [32], 21
(10 to 30) features based on angles that calculated as (2) [33],
and 13 (31 to 43) features based on rational fingers’ length.
The resultant feature vector having the size (number of images
multiplied by number of features/image) will be used.
slope = y2 - y1 (1)
x2 - x1
u·v (2) Fig. 5. Statistics of the created dataset.(a) Number of
cos ? = ?u??v? individuals per family. (b) Male-female count per family.
u and v are vectors.
were mapped to a distinct label, and the output was fed to
D. ANN Classifier fully connected layers using ReLU activation function. We
A neural network (NN) is one of the most used machine created an ANN model with two hidden layers of 10 neurons
learning algorithms [34]. The fundamental components of each. The final layer has a 14-node output and a softmax
a typical neural network are lots of small, interconnected activation function. Fig. 8 and Fig. 9 show visual mode
processing units, or ”neurons,” each of which generates a and table mode of network’s structure. The network has 444
series of real-valued activations for the intended outcome. An trainable parameters and 0 non-trainable parameters.
ANN is composed of a network of artificial neurons, also
known as network nodes [35]. The ANN structure is critical IV. RESULTS AND DISCUSSION
in solving problems, in the case that the number of input and
output is fixed, the performance of the ANN model depends Our own established dataset contains the hand images of 81
on the number of hidden layers and the neurons in each hidden people. The images were acquired using a mobile camera.
layer. The characteristics retrieved from each hand picture There were no pegs to direct the fingers, rings and other hand
ornaments could be worn without being restricted. The classi-
fication task was achieved by building the appropriate neural