Big Data Cogn. Comput.
2022, 6, 128
7 of 22
5.1. Handcrafted Models
In handcrafted models, we extract various ageing features manually. Extracting facial
features begins by using either a single filer or a combination of different filters such as
Gabor [
26
], Histogram of Oriented Gradients (HOG) [
27
], or Sobel filters [
28
]. The filter’s
parameters are manually tweaked to acquire as many features as possible. The filters are
slid over an image to extract features such as wrinkles, head shape, textures, or edges
that indicate the subject’s age. Handcrafted models generally require less computational
power as they are not as complex as their deep-learning counterparts. However, one
drawback is that these models are usually less accurate. This section analyses several
handcrafted methods used to manually extract facial features from a photo. Generally,
there are four methods:
1.
Anthropometric models
: Anthropometric measurements are quantifiable dimensions
of the bones, muscles, and the human body’s overall structure [
24
]. These measure-
ments help us understand the geometrical structure of the human body and can help
us distinguish between different age groups and genders. Several studies [
25
,
26
] used
anthropometric models We can represent the anthropometric models in Equation (1)
in which R is the circle’s radius (face) and theta is the initial angle formed with the
vertical axis. K is a variable that increases over time and R
0
represents the growth of
the human face over time.
R
0
=
R
(
1
+
k
(
1
−
cos θ
))
(1)
2.
Do'stlaringiz bilan baham: