Professional Documents
Culture Documents
8age and Gender Estimation
8age and Gender Estimation
Abstract—In this paper, we propose age and gender estimation using averaged faces of people from 25 to 60 years old used
system by vairous features. Age and gender has a lot of charac- a method of focusing on face texture and shape. Ueki et al.
teristics. These characteristics are one of the difficult cognitive [7] reported a method of age-group classification by linear
process in human interaction. If we can extract the important
feature of this cognitive process, it is considered that the age and discriminant analysis (LDA). Takimoto et al. [8] proposed a
gender estimation by the machine becomes possible. Therefore, gender and age estimation technique that is not influenced
we propose a method of age and gender feature extraction and by posture changes by estimating a NN from several features
estimation using the face image. In this paper, the age of the face including the face texture and features.
means apparent-age that is based on the human perception of However, there are some problems with almost of all
age. Moreover, person’s aging and gender difference appear in
the faces. For example, the pigmented spot, the wrinkle, sagging conventional methods. The most important problem is that
skin, shape, color of skin, and so on. Thus, we extract these these methods are only roughly classified the person’s age.
several features for age and gender estimation. Furthermore, we For example, 20 years old and 29 years old are classified into
estimate a continuous age and gender using a neural network the same class but 19 years old and 20 years old are classified
(NN). into different classes when the estimated age is classified into
groups of ten year width using conventional techniques such
I. INTRODUCTION
as those mentioned above. However, 19 years old and 20 years
In our life, we use the age and gender information in the old are different by only one year old. This difference is an
various situation. In particular, we meet the person a lot in important problem in the area of age estimation. Thus, it is
daily life and we can take a smooth and flexible response assumed that the reliability of recognition accuracy is doubtful.
routinely by estimating age and gender. For example, if we saw Moreover, it is difficult to extract a suitable feature in tech-
the elder person, we behave politely. This is unique to humans niques for which feature extraction is necessary. Furthermore,
sense. It is unknown that why the human conjecture the age most of conventional method needs to detect a lot of feature
and gender, roughly. Therefore, we pay attention to this human point exactly. If it can’t, recognition accuracy falls. In addition,
ability, the same ability as human is given to the computer, age the above studies are based on estimating the actual age, and
and gender are estimated. By the way, age estimation methods there has been little research on human age perception.
based on images of face have been widely studied [1]-[9]. Then, we proposed novel age estimation system [9]. In this
In a study on the change in the physical shape of the face method, we reported that the proposed method can estimate
with age, Todd et al. [1], [2] indicate that the contour of the the apparent-age and gender by frequency feature of a face.
skull can be approximated by a cardioid transform. Yamaguchi However, it is considered that other various features are seen
et al. [3] confirm that the differences between the features of when we estimate the age, and other study reported the
an adult’s face and a child’s face include the length of the effectiveness of a various features [8].
face and the ratio of each part. Age estimation by computer has Also, it is considered that gender information is important
also been performed. Kanno et al. [4] show that a male can be characteristics of a person. Then, gender estimation methods
identified by neural networks representing four ages (12 years, based on images of face have been widely studied, too [3],[8].
15 years, 18 years, and 22 years). Kwon and Lobo [5] reported Moreover, these methods estimate gender by using similar
that the proposed method has been implemented to classify features of the age.
input images into one of three age-groups: babies, young Therefore, we propose an apparent-age estimation system
adults, and senior adults by using the placement information based on several features. In addition, we propose gender
and texture information. However, almost of all their studies estimation system by using age features. In this study, not
were based on cranio-facial development method and skin only the texture feature of the full face but also a local
wrinkle analysis. Burt and Perrett [6] studied age perception texture feature, color feature, shape of face feature et al.
180
TABLE II
S HAPE FEATURE
Feature
Eyebrow Thickness, Length
Eye Area, Angle
Nose Width
Mouth Length, Thickness, Corner
Shape Facial contour
Distance Jaw from eyes, Nose from eyes
Mouth from eyes, Mouth from nose Fig. 3. Sample of the frequency feature
Jaw from mouth
181
Fig. 6. Extraction area of the color feature
3) Texture Feature: Pigmented spot and wrinkle on skin A. Face Image Database
area are considered as minute change noise. Then, we use the
The face database was provided from the Human and Object
ε-filter for extract this feature [12]. The ε-filter is defined as
Interaction Processing (HOIP) organization in Japan [10]. The
follows:
subject images conprised people with a wide range of ages,
∑
N
and we selected subjects that did not sport a pair of glasses.
y(n) = x(n) + ak F (x(n − k) − x(n)) (9) The background were made the same for all subjects. Subjects
k=−N
were directed to face the lens of the camera, and a picture was
|F (x)| ≤ ε : −∞ ≤ × ≤ ∞. (10) taken with a neutral expression (Table III). 252 people whose
images were preprocessed beforehand were used as subjects.
There is a difference of the I/O signal below ε when it is a In this paper, the face database was used with permission from
nonlinear function that F (x) shows in Fig.4. In this study, we Softopia corporation, Japan. It is prohibited to copy, to use,
use the ε-filter which has 7 size of window. Moreover, we and to distribute the images without the authorization of the
set ε value as 20. Example result of the ε-filter is shown in copyright holder.
Fig.5. We subtract the result image of the ε-filter from original
image and we make the histogram. We adopt these histogram
B. Apparent-age Database Creation
as texture features.
4) Color Feature: It is considered that the color information In this study, we note the human apparent-age, and the
has the correlation in the age. Then, we select the color feature apparent-age was defined as an age that a lot of person felt.
for age feature. In this study, we use the L∗ a∗ b∗ color system In this paper, the apparent-age is obtained by conducting a
that is the color system based on human perception. The color questionnaire with 58 respondents. It is considered that the
feature is extracted from the cheek area and the lip area, objectivity of the apparent-age, which people evaluate subjec-
respectively (Fig. 6). Furthermore, we define as the age feature tively, increases with the number of respondents. Moreover,
each part average and variance. the error of age estimation for each age can be reduced by
using respondents of various ages and both genders.
C. Apparent-Age and Gender Estimation In the questionnaire method, the respondent sees a face
We estimate the apparent-age and gender using a NN. We image and the apparent-age is given. The face images are
use a 3-layered NN. There is one hidden layer in this NN, and presented at random. The apparent-age is assumed to be the
the back propagation method is used for learning. In this paper, median value of an ages given by the respondents. In the
output layer is adjusted to one to enable continuous estimation proposed method, this age is adopted as the apparent-age for
of the age, and the estimated age is calculated from the output the teacher data.
182
TABLE IV
AGE ESTIMATION ERROR 80
apparent-age
Frequency feature 5.08 years 70 estimated age
Texture feature 5.69 years
Color feature 12.27 years 60
Shape feature 5.79 years
All feature 4.65 years 50
age
Human age perception 4.94 years
40
30
C. Conditions of Apparent-Age Estimation 20
In order to show the effectiveness of the proposed method, 10
we perform a simulation. In this paper, we use the actual data 0 20 40 60 80 100 120
that is provided from HOIP organization in JAPAN [10], and subject number
we use a sample size of only 113 men. Moreover, the feature
point and color area is extracted manually. Furthermore, we Fig. 7. Apparent-age and estimated age of all subjects
selected subjects that did not sport a pair of glasses. In this
simulation, we use the leave-one-out cross-validation method.
18
D. Apparent-Age Estimation Results and Discussions 17 proposed method
16 texture
Table IV shows the average estimated age error of each 15
14 freq
183
7.5
7 proposed method male
6.5 female
5.5
5 80
4.5
4 60
3.5
3
2.5 40
2
1.5 20
1
0.5
0 0
10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80
generation generation
Fig. 9. Age estimation error of each generation (proposed method) Fig. 11. Recognition accuracy of gender estimation (shape feature, each
generation)
100
In the future works, it is considered that the improvement
recognition accuracy (%)
184