Hindawi

Computational Intelligence and Neuroscience

Volume 2021, Article ID 8456736, 23 pages
https://doi.org/10.1155/2021/8456736

Research Article
Cognitive Matching of Design Subjects in Product Form
Evolutionary Design

Shutao Zhang , Shijie Wang , Aimin Zhou , Shifeng Liu , and Jianning Su
School of Design Art, Lanzhou University of Technology, Lanzhou 730050, China

Correspondence should be addressed to Aimin Zhou; [email protected]

Received 30 May 2021; Accepted 18 September 2021; Published 1 October 2021

Academic Editor: Daniele Bibbo

Copyright © 2021 Shutao Zhang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

In actual product development, the cognitive differences between users and designers make it difficult for the designed products to
be recognized by users. To reduce the cognitive differences between these two design subjects, this paper proposes a method of
cognitive matching of the design subjects. First, we use the relevant methods of Kansei engineering to quantify the Kansei image
cognition of the two design subjects and construct a cognitive matching model of the design subjects with information entropy
and the technique for order preference by similarity to ideal solution (TOPSIS). Second, according to the Kansei image, the Kansei
image prototype cluster is constructed, and the representative Kansei image prototype is obtained. Then, we combine an artificial
neural network (ANN) with a cognitive matching model of the design subjects to construct a product Kansei image evaluation
system; this is used to evaluate the evolved forms. Finally, a product Kansei image form evolution system is constructed based on
the genetic algorithm (GA). To some extent, the system simulates the cognitive matching process between designers and users in
product design, helps designers to more accurately understand the cognitive trends of the two design subjects, and provides a
theoretical basis for the intelligent design of product forms through the cognitive balance of multiple design subjects. This paper
takes a beverage bottle as an example to verify the feasibility of the model through a comparative study.

1. Introduction very large differences between designers and users in terms

of knowledge, educational background, personality char-
Design is a creative process of expressing human needs in acteristics, and understanding of the product. Therefore, by
concrete product forms, and its main purpose is to endow extracting Kansei variables related to product design ele-
things with new value [1]. As a medium of the expression of ments, we can transform users’ emotional characteristics
designers’ ideas and multilevel symbolic meaning, the creative into product attributes and accurately grasp and understand
product form is the optimal solution for expressing users’ the Kansei image cognition trends between users and de-
subjective needs through the characteristics of shape, function, signers, which can help designers construct product forms
and material. With the development of science and technology, that meet users’ needs [3]. For example, K. Qiu et al. used
product development technology has gradually transformed Theil entropy to evaluate the cognitive friction between users
from traditional design and development methods to intelli- and designers and built a cognitive friction balance model to
gent design modes based on computer technology, and the alleviate the cognitive friction between the two groups [4].
focus of product design has evolved from functional re- Based on double coding in psychology, Hu et al. decom-
quirements to user-centred emotional satisfaction. posed the cognition of automobile modelling Kansei images
In product design, designers, users, and products are the to compare the differences between users and designers in
three most important aspects. Designers and users are the semantic word selection and corresponding morphological
core subjects of design. Designers infuse their own cognition features [5].
into the product form through communication with users to In the traditional design process, designers use sketches
meet the emotional needs of users [2]. However, there are to express the product scheme, which is tedious and requires
2 Computational Intelligence and Neuroscience

strong drawing ability. The emergence of intelligent design there are studies on the Kansei image cognition of different
provides new theories and methods for product form design design subjects, most of them focus on comparing the
which greatly change the task that designers undertake in the differences in the product Kansei image cognition among
design process and provide a new way to shorten the design different design subjects, while there are relatively few
cycle. For example, Hsiao and Tsai constructed an automatic studies on the quantitative description and matching in the
design system using a fuzzy neural network and genetic Kansei image cognition of multiple design subjects. In ad-
algorithm (GA) and made a preliminary attempt at intel- dition, in product evolutionary design, there are few studies
ligent product design [6]. As the core of intelligent design, on Kansei image traction for optimizing product form.
evolutionary algorithms have been widely used in product Therefore, integrating user cognition into product form and
design and have become one of the main means of product accurately grasping the cognitive relationship between de-
form evolution and evaluation [7]. Common evolutionary sign subjects to realize the cognitive matching of design
algorithms usually take the existing optimal design scheme subjects represent an important research direction.
as the parent sample for the genetic operation and generate a Based on the above problems, this paper proposes a
large number of new schemes. This method rapidly produces method of the cognitive matching of design subjects in
a large number of design schemes and also produces new product form evolutionary design and explores the matching
problems; that is, with increasing evolutionary generation, relationship of Kansei image cognition between users and
the Kansei images of the product forms of many offspring designers in the process of design scheme generation in a
gradually deviate from the target Kansei image. In reality, the quantitative way. First, we apply the methods of Kansei
Kansei image cognition of human beings has corresponding engineering to explore the inherent trend of Kansei image
product form characteristics. We concretely express the cognition between users and designers and construct the
Kansei image information of consumers to produce the cognitive matching model of design subjects. Second, we
product form representing the target Kansei image and call it introduce product prototype theory and the Kansei image
the Kansei image prototype [8]. Combining the Kansei cognitive dynamic mechanism and establish a Kansei image
image prototype and the design scheme that needs to evolve, prototype cluster. Next, we determine a representative
and with the offspring inheriting the characteristics of the Kansei image prototype and take it as external traction to
parents, we take the morphological elements of the Kansei lead product form evolution. Then, the product Kansei
image prototype to lead the evolution of the offspring image evaluation system is constructed by combining an
product form to help ensure that the offspring product form artificial neural network (ANN) with the cognitive matching
has the characteristics of the target Kansei image and always model of design subjects to evaluate whether the product
evolves towards the direction of the target Kansei image. We evolution form has reached a state of a strong match in the
call this process Kansei image traction. Kansei image cognition of the two design subjects. Finally,
In recent years, many scholars have conducted research we build a product Kansei image evolution system based on
on the Kansei image cognition of users and designers. For a GA and verify the feasibility of the model by comparing the
example, to study the Kansei image needs and personality results. This study provides a quantitative method and new
preferences of users, Yan et al. used the semantic difference research ideas for Kansei image cognitive matching between
(SD) method and language variables to obtain the Kansei users and designers and has guiding significance for the
data of products to obtain the satisfaction in product continuation of the product Kansei image style in the
evaluation [9]. Reference [10] used eye tracking technology process of product form evolution.
to collect users’ needs and obtained users’ Kansei image The rest of this article is structured as follows. Section 2
cognition and objective evaluation data of products by introduces the related theories such as cognitive differences,
precise methods to meet the needs of specific target users. cognitive matching, and the technique for order preference by
Reference [11] constructed a cognitive model to explore the similarity to ideal solution (TOPSIS). Section 3 introduces the
relationships between design subjects, the design process, basic process of the product form evolution method based on
and the cognitive process. Xue et al. constructed an inte- the cognitive matching of the design subjects. Section 4
grated decision-making system of optimized product Kansei demonstrates the above process with a case. Section 5 presents
image design by using the methods of quantification-I analysis and discussion of the results and shortcomings of this
theory and grey correlation analysis to simulate the Kansei study. Finally, Section 6 is a summary of this paper and
image evaluation of products by users to optimize the explains the significance of the model.
product form [12]. To better retain the characteristics of
target Kansei images in product families, Zhang et al. 2. Related Theories
proposed a multicriteria decision system of products from
Kansei images based on a logistic regression model [13]. Xu 2.1. Cognitive Differences. Cognition is an information
et al. studied the differences in product Kansei image processing operation in which humans recognize objects in
cognition between users and designers by using the method the environment through their senses and store the acquired
of gene network comparison, which helped designers more object information in a structured form [15]. As the media of
accurately understand the implicit needs of users and im- designers’ cognitive expression, products communicate with
prove the market recognition of products [14]. users through their form, function, structure, and other
At present, the related research mainly explores the aspects to convey designers’ ideas. According to Maslow’s
Kansei image cognition of a single design subject. Although hierarchy of needs theory, when people’s basic needs are
Computational Intelligence and Neuroscience 3

met, they pursue higher levels of needs [16]. With the gradually strengthen. Under the stimulation of the cognitive
continuous advancement of science and technology, emo- system and external environment, users can perceive the
tional needs have gradually become the main factor affecting connotation conveyed by a product when they contact the
consumers’ purchase of products. Lee and Chen used in- product. Even if they have never contacted the product form,
teractive qualitative analysis (IQA) to conduct psychological users can generate cognitive structure according to their own
model analysis on different groups and explored the cog- cognitive memories, thus connecting the form elements with
nitive differences between different groups when using symbolic meaning [22]. In the actual production process,
smartphones [17]. Reference [18] discussed the cognitive due to the differences in social responsibilities between users
process of users on products and the differences between and designers, the two design subjects seldom directly
designers and users’ mental models and proved that only communicate, meaning the product form is unable to reflect
products that conform to users’ mental models can meet cognitive consistency between users and designers. To
users’ needs. Due to the influence of background factors, minimize cognitive differences, designers need to grasp the
there are very large differences in the understanding of symbolic meaning perceived by users when they contact a
products of users and designers. That is, users experience product. According to the cognitive mode, designers con-
imperfect cognition of their own class, personality, and struct a product form recognized by users and satisfy users’
knowledge and thus different understanding of the scenes psychological expectations from the visual system to realize
needed by the products. Their cognition of products forms cognitive matching. Therefore, we can use computer tech-
according to their own experience and knowledge. However, nology for cognitive simulation to fully understand the
designers obtain a view of products from the aspects of cognitive processes of users and designers [23].
demand mining, product positioning, product innovation At present, cognitive matching is widely used in other
and rationality, design expression, and other aspects. The fields. In the field of medicine, Zane et al. found cognitive
two design subjects do not directly communicate, resulting differences between outpatients and clinicians in treatment
in differences in the cognitive perception of the product, as and found that cognitive matching between them can affect
shown in Figure 1. the course of treatment and predict the treatment outcome
The development of the market economy has promoted [24]. To improve the inaccuracies in traditional clinical
the gradual development of product design from meeting reasoning and the main cognitive biases in medicine, [25]
the needs of the public to customized experiences. The status established a cognitive balance model by using the diag-
of users in design has also been transformed from the re- nostic process of a fuzzy cognitive map. In the field of
cipients of products to participants in product design. psychology, Roth et al. established a reclassification model
Therefore, the cognitive differences between users and de- based on cognitive consistency and explored whether the
signers have become the focus of many scholars. For ex- identity between individuals and groups determines the
ample, Yao and Huang abstracted multiple shape compatibility between different groups [26]. In the field of
information into simple symbols based on the concept of management, Hassan and Ralf studied human cognition and
abstract symbols and explored user cognition by establishing behaviour through the analysis of cognitive matching and
a mapping relationship between user cognition concepts and predicted and discussed the use of accounting information
product shape information [19]. To prevent the failure of the [27]. In the field of design, to study the perception differ-
user experience due to cognitive friction, He qualitatively ences between users and designers in product modelling,
analysed the causes and avoidance methods of cognitive [28] constructed a Kansei image matching model between
friction in the process of product design and use [20]. Based users and designers. Su et al. developed a cobweb colliding
on the clear perception characteristics of users and de- evolutionary system of product form to study the cognitive
signers, [21] proposed a model of a team style-promoting balance between users and designers by simulating the
design scheme in the face of perception conflict to explore design thinking of the cobweb structure and spider pre-
the influence of the design team’s perceived conflict be- dation mode [29].
haviour style on the scheme process and product in inter- In summary, research on cognitive matching in different
action design. disciplines can solve the problems caused by cognitive bias
The existing studies mostly focus on the qualitative to a certain extent. The related research in the field of design
exploration of the differences, factors, and influences be- mainly focuses on exploring the cognitive matching of the
tween the two design subjects, while there are relatively few two design subjects from the qualitative level, and there are
studies on the cognitive trends of users and designers in the relatively few quantitative studies on the cognitive trends of
process of product scheme generation. Therefore, this paper design subjects in the iterative process of a single scheme.
uses the quantitative method to study the trends of Kansei Therefore, we introduce the cognitive matching mechanism
image cognition of the two design subjects in the process of into the product design process to quantitatively explore the
product scheme generation. cognitive variation in users and designers during product
scheme iteration and propose a cognitive matching model.

2.2. Cognitive Matching. The user’s cognition of a product

mainly depends on the users’ perception and the interaction 2.3. TOPSIS Method. TOPSIS is a multicriteria decision
process between them. With the continuous enrichment of analysis method that was developed by Ching-Lai Hwang
the users’ experience, their cognitive memories of products and Yoon in 1981. The basic idea of this technique is to
4 Computational Intelligence and Neuroscience

This water bottle is

mainly curved, with a The shape of this water
concave center for easy bottle is very nice.
grasping.

product
Designer User
Figure 1: Map of cognitive differences.

determine the optimal scheme (positive ideal solution) and Kansei image of users and designers to obtain the coordi-
the worst scheme (negative ideal solution) according to the nation index of the two design subjects when they reach
original scheme and to compare the distances between all the cognitive matching and to calculate the cognitive matching
original schemes and the optimal scheme and the worst degree between the two design subjects.
scheme to judge the pros and cons of each scheme. The
determined alternative scheme should be located at the 3. Cognitive Matching of Design Subjects in
shortest geometric distance from the optimal scheme and
Product Form Evolutionary Design
the longest geometric distance from the worst scheme [30].
As a multicriteria decision analysis method, TOPSIS has The overall research process is divided into four parts,
been successfully applied to industrial technology [31, 32], namely, the cognitive matching model of design subjects,
computer science [33–35], medical problems [36], applied determination of the Kansei image prototype, the product
statistics [37–40], environmental protection [41, 42], and Kansei image evaluation system, and the product Kansei
other fields. image form evolution system, as shown in Figure 2. First, we
In the field of design, to address a design that is infeasible investigate the Kansei images of users and designers with
due to the uncertainty in customer preference in the cog- representative samples and Kansei images and establish the
nition of the optimization goals, Zhou et al. used TOPSIS to standardized matrix of the two design subjects. Based on the
rank factors such as production cost, time, product output, entropy weight method and TOPSIS and other methods, we
and pollution emissions [43]. Based on entropy theory and construct the cognitive matching model of the design
TOPSIS, Wang et al. proposed a comprehensive evaluation subjects and determine the cognitive probability of the two
system with subjective and objective combinations to im- design subjects by averaging the standardized matrix. Sec-
prove ergonomics evaluations and decision-making in ond, according to the selected Kansei image of user research,
product development [44]. To determine the best conceptual we obtain the product prototype representing the Kansei
design in the process of concept evaluation, [45] introduced image. After screening, we construct the Kansei image
the customer evaluation and design specifications described prototype cluster and select the product prototype with the
by soft foundation theory and entropy theory into the highest degree of recognition as the representative Kansei
TOPSIS framework to rank the pros and cons. On the basis image prototype. Then, according to the corresponding
of establishing the relationship between consumer demand relationship between the sample data and the Kansei image
and technical characteristics, [46] calculated the basic im- evaluation data, the Kansei image evaluation system of users
portance of technical characteristics and used the TOPSIS and designers is constructed by using an ANN. Combined
multiattribute decision model to modify the basic impor- with the cognitive matching model of the design subjects, the
tance of technical characteristics to overcome the limitation product Kansei image evaluation system is constructed.
of considering only consumer factors. Su et al. used the Finally, we take the product prototype and representative
TOPSIS method to prioritize new samples coupled with Kansei image prototype as the parents for genetic operation.
different weight design elements to explore the coupling Through the judgement of the fitness function, we select the
characteristics of elements in product Kansei image product form with the highest matching degree to replace
design [47]. the product prototype, repeat the genetic operation, and
TOPSIS is usually used to analyse the pros and cons of finally select the product form with high matching cognition.
multiple indicators to select alternative indicators. In this We conduct a cognitive investigation on the two design
study, the TOPSIS method is used to calculate the relative subjects to verify the feasibility of the model through
closeness between the Kansei image cognition and the target comparative analysis.
Computational Intelligence and Neuroscience 5

Determination of Target Product

Kansei Image Prototype
Cluster Product Prototype and Kansei
Image Prototype
Collection of Kansei Image Collection of Product Samples
Replacement

Product Kansei Image Set Product Sample Set Representative Kansei Data Initialization
Image Prototype
Cluster Analysis Product Form with The
Confirmation of Kansei Highest Matching Degree
Representative Product Image Prototype
Representative Kansei Images
Samples
Genetic Operation
Semantic
Difference Method Parameterization of
Research on Users and Contour
Designers

Choosing a Kansei Image

The Emergence of New Forms
Normalization
Obtaining Key Point
Standardized Matrix of Coordinates Ranking of Cognitive
Cognitive Subjects Matching
Average Entropy
Processing Method Training
Whether to reach
The Determination of Kansei Kansei Image Weight of equilibrium
Image Cognitive Probability Cognitive Subjects Neural Networks For Users No
and Designers
Yes
Weighted Evaluation of
Cognitive Subjects
Output of Image Form
Determination of Mutation
Probability and Crossover
Probabilty Semantic Difference
Method
Cognitive Matching Model of Product Kansei Image
Design Subjects Evaluation System Method Validation

Product Kansei Image

Cognitive Matching Model of Design Subjects Product Kansei Image Form Evolution System
Evaluation System

Figure 2: Research process.

3.1. Cognitive Matching Model of the Design Subjects a11 · · · a1l

⎡⎢⎢⎢ ⎥⎥⎤⎥
3.1.1. Construction of the Evaluation Matrix of the Design A � ⎢⎢⎢⎣ ⋮ ⋱ ⋮ ⎥⎥⎥⎥⎦,
⎢ (1)
Subjects. al1 · · · all
(1) Determination of the Representative Product Samples.
We collect pictures of product samples from jour- where A is the similarity evaluation matrix of the
nals, websites, and other channels. To prevent visual Kansei words. all is the similarity evaluation value.
interference from factors such as colours and pat- (3) Obtaining the Evaluation Matrix. Combined with
terns, we extract sample contours to establish the representative samples and Kansei images, the SD
product sample set and determine representative questionnaire is designed to evaluate Kansei image
samples by expert interviews. cognition. The commonly used SD questionnaire is
(2) Determination of the Representative Product Kansei divided into a 5-level scale and a 7-level scale. In this
Image. The Kansei words of the target product are paper, the SD questionnaire of the 5-level scale is
collected from journals, websites, and other chan- used to investigate the Kansei image cognition of the
nels. We can obtain the most representative Kansei two design subjects, and the Kansei image matrix is
words of the target product Kansei image from constructed. The results are as follows:
expert interviews and construct the Kansei image set;
namely, S � [s1, s2, . . ., sl], where S is the Kansei image φ φ · · · r1j
φ
set and sl is the Kansei image. ⎡⎢⎢⎢ r11 r12 ⎤⎥⎥
⎢⎢⎢ φ φ ⎥ ⎥
We use semantic similarity to investigate the Kansei ⎢⎢ r21 φ
r22 · · · r1j ⎥⎥⎥⎥⎥
Rφ � ⎢⎢⎢⎢⎢ ⎥⎥⎥, (2)
word similarity of the experts, to establish the similarity ⎢⎢⎢⎢ ⋮ ⋮ rφxy ⋮ ⎥⎥⎥⎥⎥
evaluation matrix shown in formula (1), and to cluster ⎢⎣ φ ⎥⎦
φ φ
the results to obtain the representative Kansei images. ri1 ri2 · · · rij
6 Computational Intelligence and Neuroscience

where Rφ is the Kansei image recognition evaluation provides a theoretical basis for the practical application of
matrix. φ represents the design subject, and φ � 1 and entropy theory. For example, Wan et al. [49–51] proposed
φ
2 represent users and designers, respectively. rij is fuzzy entropy, hesitant entropy, and cross-entropy to de-
the evaluation value of the φ-th design subject on the termine the weights in multiattribute group decision-
j-th Kansei image of the i-th sample. making problems. Xu et al. [52] proposed the fuzzy entropy
and hesitant entropy of probabilistic linguistic term sets
The b-th Kansei image is selected from the representative
(PLTSs) to calculate the total entropy of PLTSs and used the
product Kansei images for further research, and the eval-
cross-entropy to determine the attribute weight of the de-
uation matrix for the b-th Kansei image is obtained as
cision system.
follows:
In recent years, information entropy has been intro-
φ φ φ φ T duced into the design field to explore the entropy values of
rb � 􏽨r1b r2b · · · rib 􏽩 , (3)
various indicators of products [53]. This paper uses infor-
φ
where rb is the Kansei image cognitive evaluation matrix for mation entropy to analyse the Kansei image cognition of
the b-th Kansei image. φ represents the design subject, and users and designers to determine the weight relationship
φ
φ � 1 and 2 represent users and designers, respectively. rib is between these two design subjects. The calculation formula
the evaluation value of the φ-th design subject of the i-th of this entropy is as follows:
sample of the b-th Kansei image. m
To reduce the errors generated by the two design subjects Qφ � −k 􏽘 Pφi ln Pφi , (6)
in the evaluation process, we use formula (4) to normalize i�1
the cognitive evaluation matrix and to obtain the decision φ
where Qφ is the entropy value of the φ-th design subject. Pi
matrix V.
represents the proportion of the Kansei image cognitive
φ φ
φ rib − min rib probability of the i-th sample of the φ-th design subject in
vib � φ φ, (4) the Kansei image cognitive probability of the cognitive
max rib − min rib φ
subject, and 0 < Pi < 1. k is a constant, 1/ln m.
φ
where vib is the normalized evaluation value of the φ-th
(1) Formula (7) is used to obtain the proportion of the
design subject of the i-th sample of the b-th Kansei image.
Kansei image cognition probability of each sample in
the Kansei image cognition probability of the design
3.1.2. Determination of the Cognitive Probability. The de- subject.
cision matrix V is averaged to obtain the probability matrix φ
φ vib
Eφ of the Kansei image cognition between users and Pi � m φ. (7)
designers. 􏽐i�1 vib
φ
Eφ � 􏽨eb 􏽩, (5) φ
(2) We substitute Pi into formula (6) to obtain the
φ entropy value of the φ-th design subject, and the
where eb is the cognitive probability of the φ-th design weight Wφ of the design subject in the evaluation
subject of the b-th Kansei image. process is
Because designers have received systematic design
knowledge training, their cognition of products is mainly 1 − Qφ
Wφ � , (8)
based on the feasibility of product solutions, and their 􏽐nφ�1 1 − Qφ 􏼁
cognition of products represents a small jump. However, due
to the influence of complex factors such as educational where n � 2.
background, users usually form their opinions based on their (3) The Kansei image cognitive evaluation of users and
own needs, and there is a great leap in their cognition of a designers is weighted to obtain the weighted cog-
product. Therefore, the Kansei image cognitive probability nitive evaluation value Zφ of each design subject:
of two design subjects is regarded by means of the crossover
φ
probability and the mutation probability to guide the sub- Zφ � Wφ rb . (9)
sequent evolution process of the product Kansei image form.
(4) TOPSIS is introduced to calculate the relative
closeness between the Kansei image cognition of the
3.1.3. Construction of the Cognitive Matching Model of the
design subject and the target Kansei image; namely,
Design Subjects. Information entropy, proposed by Shan-
non in 1948, is a measure of the diversity or uniformity of dφ−
microstates in thermodynamics. By calculating the entropy tφ � , (10)
dφ− + dφ+
of the system, the probability of the system in this state can
􏽶������������
be judged; that is, the greater the entropy value is, the greater 􏽴
m
the probability is. The smaller the entropy value is, the φ φ+ 2
dφ+ � 􏽘 􏼐vib − vb 􏼑 , (11)
smaller the probability is [48]. Some scholars have con- i�1
ducted in-depth research on information entropy, which
Computational Intelligence and Neuroscience 7

􏽶������������ (6) Combined with the idea of game theory, the un-
􏽴
m
φ− φ φ− 2 determined coefficient aφ is optimized to minimize
d � 􏽘 􏼐vib − vb 􏼑 , (12)
the deviation between Zφ and Zβ as follows:
i�1
n

φ+
min 􏽘 aφ ZφT − Zβ , (β � 1, . . . , n). (15)
where d represents the Euclidean distance between φ�1
the cognitive evaluation of the φ-th design subject
and the cognitive evaluation of the optimal scheme. According to the differential properties of the matrix,
dφ− represents the Euclidean distance between the the optimal conditions of formula (15) are obtained
cognitive evaluation of the φ-th design subject and as follows:
φ+
the cognitive evaluation of the worst scheme. vb is n
expressed as the cognitive evaluation value of the 􏽘 aφ Zβ ZφT � Zβ ZβT . (16)
optimal scheme of the φ-th design subject for the b- φ�1
φ−
th Kansei image. vb is expressed as the cognitive
evaluation value of the worst solution scheme of the Based on the constraint conditions of combination
φ-th design subject of the b-th Kansei image. determining weights on maximizing deviations, the
Thus, the coordination index C of the Kansei image improved optimization model is determined:
cognition of the two design subjects is obtained: n 􏼌􏼌􏼌 􏼌􏼌
􏼌
􏼌􏼌 n φ β φT β βT 􏼌􏼌
min f � 􏽘 􏼌
􏼠 􏽐 a Z Z 􏼡 − Z Z 􏼌􏼌,
􏽑nφ�1 tφ a1 ,a2 􏼌
􏼌 􏼌􏼌
C� β�1 􏼌 φ�1
2 (13) (17)
􏼐􏽐nφ�1 tφ 􏼑 . n
2
s.t aφ > 0 φ � 1, . . . , n 􏽘 aφ 􏼁 � 1.
φ�1
(5) The evaluation model of the comprehensive cogni-
tion coordination is constructed as follows:
The Lagrange function is established to solve the
n model as follows:
T� 􏽘a Z . φ φ
(14)
φ�1

􏼌􏼌 􏼌􏼌
􏼌􏼌 n
n 􏼌􏼌 n
L aφ , λ􏼁 � 􏽘 􏼌􏼌􏼌⎛
􏼌 ⎠ − Zβ ZβT 􏼌􏼌􏼌 + λ ⎛
⎝ 􏽘 aφ Zβ ZφT ⎞ ⎝ 􏽘 aφ 􏼁2 − 1⎞
⎠. (18)
􏼌 􏼌􏼌 2
β�1􏼌 φ�1 􏼌 φ�1

Thus, when the cognition of the two design subjects 3.2. Determination of the Kansei Image Prototype. To make
is closely matched, the solution of the undetermined the product evolution operation convenient, based on the
coefficient is as follows: results of the Kansei image survey of users in 3.1.1, we take
the average value of the survey results as the judgement
􏽐nβ�1 Zβ ZφT standard; that is, when rib ≥ rib , we select the sample to
aφ � 􏽱�����������������2 . (19)
􏽐nφ�1 􏼐􏽐nβ�1 Zβ ZφT 􏼑 construct the Kansei image prototype cluster, U � [u1, u2, . . .,
uh], where uh is a Kansei image prototype. The evaluation
value of the Kansei image cognition can reflect the repre-
The undetermined coefficient is normalized to obtain
sentative degree of product samples to Kansei images; that is,
the optimized undetermined coefficient:
the higher the evaluation value is, the higher the repre-
aφ sentative degree is, and the higher the user’s Kansei image
aφ∗ � . (20)
􏽐nφ�1 aφ recognition is. Therefore, the Kansei image prototype with
the highest cognitive evaluation value is selected as the
(7) We substitute the optimized undetermined coeffi- representative Kansei image prototype.
cient into formula (14) and obtain the cognitive
matching model of the design subjects as follows:
√��� 3.3. Product Kansei Image Evaluation System. Based on the
D � CT. (21) sample set, we parameterize the sample contour, train the
ANN with the Kansei image survey results, and finally es-
When the cognitive matching degree D ≥ 0.75, the tablish the product Kansei image evaluation system with the
cognition of the two design subjects reaches the cognitive matching model of design subjects. The specific
matching state. process is as follows:
8 Computational Intelligence and Neuroscience

Step 1: The contours of all samples in the sample set 4.1. Kansei Image Cognition Experiment
established in 3.1.1 are extracted and parameterized to
obtain the key point coordinates. 4.1.1. Determination of the Representative Samples. A total
of 130 pictures of beverage bottles are collected from
Step 2: The coordinates of the key points of the sample
websites, journals, and other channels. After preliminary
are taken as the input, and the Kansei image cognitive
screening, we obtain 77 sample pictures. To prevent the
evaluation values of users and designers are taken as the
influence of colour, pattern, and other factors on the Kansei
output. The ANN is trained to obtain an ANN based on
image cognition of the designers and users, we extract and
users’ cognition and an ANN based on designers’
draw the contour lines of all beverage bottles and convert
cognition.
them into 10 × 15 cm cards, as shown in Figure 3.
Step 3: The two trained ANNs obtained in Step 2 are We invite 5 experts to screen the beverage bottle samples
combined with the cognitive matching model of the based on morphological differences and obtain 65 repre-
design subjects constructed in 3.1.3 to obtain the sentative samples, some of which are shown in Table 1 (see
product Kansei image evaluation system, which is used Table S1 in the Supplementary Material for all representative
as the fitness function of the product evolution system. samples).

3.4. Product Kansei Image Form Evolution System. The 4.1.2. Determination of the Representative Kansei Words.
product Kansei image form evolution system established in According to journals, consumer evaluation, and other
this paper is a man-machine interface interaction system channels, we collect 46 Kansei image words about the forms
based on GA and MATLAB programming. According to the of beverage bottles and select 15 Kansei words with the KJ
genetic law, we perform genetic operations on the param- method to construct the Kansei image set. We conduct a
eterized parent samples and evaluate the new form through a semantic similarity survey with 20 experts and import the
fitness function to obtain a product form with high cognitive average value of the survey results into SPSS software for
matching. The specific process is as follows: cluster analysis. The clustering results of the Kansei words are
shown in Table 2. We select the Kansei image words nearest
Step 1: A sample is randomly selected from the sample
to the centre as the representative Kansei images, including
set to be the product prototype. We input this sample
simplism, comfortable, streamlined, exquisite, and novel.
and the representative Kansei image prototype deter-
mined in 3.2 into the system as the parents for chro-
mosome coding. 4.1.3. Survey of the Design Subjects’ Kansei Image Cognition
Step 2: The crossover probability and mutation prob- and Result Statistics. We create an SD questionnaire with 65
ability determined in 3.1.2 are the input, and a genetic representative samples and 5 representative Kansei words
operation is carried out to obtain the new product and investigate 54 users and 46 designers. There are 47 and
form. 44 valid questionnaires from each group, respectively. The
Step 3: The product form generated in step 2 is es- survey data are analysed, and some results are shown in
timated by the fitness function. If the new form meets Tables 3 and 4 (see Tables S3 and S4 in the Supplementary
the requirements, the product form is output. If the Material for the survey results of the Kansei image cognition
new form does not meet the requirements, all forms of all users and designers).
are sorted according to the degree of cognitive As the research process of each Kansei image is the same,
matching, from which the product form with the this paper takes “Exquisite” as an example for verifying the
highest degree of cognitive matching is selected to method. Some of the cognitive evaluations of users and
replace the product prototype in Step 1, and Steps designers based on this Kansei image are shown in Table 5
1∼3 are repeated until a product form meeting the (see Table S5 in the Supplementary Material for all evalu-
requirements of the fitness function is obtained to ation values of “Exquisite”).
ensure that the product form is always subject to Formula (4) is used to normalize the cognitive evaluation
Kansei image traction. of the two design subjects, and the decision matrix V is ob-
tained. Some of the results are shown in Table 6 (see Table S6 in
Step 4: After conducting the Kansei image cognition
the Supplementary Material for all normalization results).
survey of the design subjects on the output product
After averaging the data in Table 6, we can obtain the
form, we compare and analyse the results with the
Kansei image cognition probability matrix of users and
results of the product Kansei image evaluation system
designers, as shown in Table 7, to determine the mutation
to verify the feasibility of the method proposed in this
probability and crossover probability.
paper.

4. Case Validation 4.2. Cognitive Matching Model Based on the Design Subjects.
To build the cognitive matching model based on the design
To prove the feasibility of the research method proposed in subjects, we use formulas (6)∼(9) to calculate the entropy,
this paper, we choose a beverage bottle as an example for weight, and weighted cognitive evaluation of the two design
verification. subjects, as shown in Table 8.
Computational Intelligence and Neuroscience 9

Figure 3: Sample cards.

Table 1: Some representative samples.

1 2 3 4 5 6 7

8 9 10 11 12 13 14

...
59 60 61 62 63 64 65

We introduce TOPSIS and obtain the relative closeness based on the design subjects and the comprehensive cog-
between the user and the target Kansei image and between nitive evaluation model T. To make the evaluation results of
the designer and the target Kansei image according to the cognitive matching model more scientific and reason-
formulas (10)∼(12), as shown in Table 9. able, only if the cognitive evaluation of the two cognitive
The relative closeness of the two design subjects in subjects is closest to the target Kansei image evaluation can
Table 9 is introduced into formula (13), and the coordination the constructed evaluation model of the comprehensive
index C � 0.25 of the Kansei image cognition of the two cognition coordination express the Kansei image cognitive
design subjects is obtained. evaluation of any cognitive subject. Therefore, based on
According to formula (21), there is a strong correlation formulas (15)∼(20), we can obtain the optimized undeter-
φ
between the evaluation of the cognitive matching model D mined coefficient a∗ , as shown in Table 10.
10 Computational Intelligence and Neuroscience

Table 2: Clustering results of the Kansei images.

Number Kansei image Category Distance
1 Simplism 1 0.33000
2 Lightweight 1 0.33000
3 Streamlined 3 0.30507
4 Textural 4 0.37350
5 Beautiful 3 0.51592
6 Comfortable 2 0.33439
7 Interesting 5 0.33864
8 Rounded 2 0.45008
9 Sporty 3 0.49180
10 Coordinated 2 0.41757
11 Individual 5 0.29168
12 Friendly 2 0.35619
13 Novel 5 0.24518
14 Exquisite 4 0.37350
15 Fashionable 3 0.44127

Table 3: Survey results of the Kansei image cognition of some users.

Sample Simplism Comfortable Streamlined Exquisite Novel
1 4.68 4.09 3.21 3.26 1.91
2 3.13 2.57 3.21 2.81 2.83
3 4.02 3.15 3.7 3.36 3.26
4 4.19 3.32 2.72 3.32 2.64
5 4.64 3.85 3.81 3.51 2.4
6 4.13 3.83 3.94 3.49 2.19
7 3.23 2.85 2.62 3.13 2.91
8 3.26 3.04 2.96 3.21 3.06
9 2.19 2.47 2.87 2.87 3.96
10 3.51 3.3 2.79 3.17 2.64
...
61 4.04 3.55 3.72 3.62 3.19
62 3.98 2.96 3.32 3.26 4.09
63 2.13 2.23 2.47 2.89 4.02
64 2.49 2.4 2.89 2.94 3.57
65 3.7 3.32 3.32 3.32 3

Table 4: Survey results of the Kansei image cognition of some designers.

Sample Simplism Comfortable Streamlined Exquisite Novel
1 4.32 3.57 2.82 3.02 1.61
2 2.75 2.98 2.91 2.95 3.36
3 3.66 3.34 4.34 3.52 3.61
4 4.61 3.36 2.2 3.14 2.75
5 4.48 3.73 3.91 3.3 2.23
6 3.91 3.75 4.2 3.32 2
7 3.18 2.84 1.89 3.02 3.14
8 2.8 2.91 3.11 3 2.77
9 1.8 2.27 2.98 3.09 4.05
10 3.3 3.09 2.41 2.82 2.86
...
61 3.8 3.39 3.23 3.27 3.16
62 4.14 3.14 3.07 3.34 4.02
63 1.91 2.16 1.75 2.7 4.2
64 1.7 2.3 2.89 2.95 3.73
65 3.36 3.18 3.36 2.89 2.75
Computational Intelligence and Neuroscience 11

Table 5: Some evaluation values of “Exquisite.” Table 9: The relative closeness.

Sample User Designer User Designer
1 3.26 3.02 d+ 14.88 15.65
2 2.81 2.95 d− 17.60 16.91
3 3.36 3.52 tφ 0.54 0.52
4 3.32 3.14
5 3.51 3.3
6 3.49 3.32 Table 10: The undetermined coefficient of each design subject after
7 3.13 3.02 evolution.
8 3.21 3
9 2.87 3.09 User Designer
φ
10 3.17 2.82 a∗ 0.59 0.41
...
61 3.62 3.27
4.3. Determination of the Representative Kansei Image
62 3.26 3.34
63 2.89 2.7 Prototype. According to Table 7 in 4.1.3, the average value of
64 2.94 2.95 the Kansei image cognitive evaluation of all samples is 0.378.
65 3.32 2.89 We compare the Kansei image cognitive evaluation of users
in Table 6 with the average value of the Kansei image
cognitive evaluation and construct a Kansei image prototype
Table 6: Some of the normalization results. cluster for all product forms whose cognitive evaluation is
higher than the average value of Kansei image cognitive
Sample User Designer evaluation to obtain Kansei image prototype cluster B. The
1 0.454 0.331 form and number are shown in Table 11.
2 0.076 0.288 Among them, the users’ Kansei image cognition eval-
3 0.538 0.638 uation value of sample 13 is the highest, so we choose sample
4 0.504 0.405
13 as the representative Kansei image prototype, as shown in
5 0.664 0.503
6 0.647 0.515 Figure 4.
7 0.345 0.331
8 0.412 0.319 4.4. Product Kansei Image Form Evaluation System. The
9 0.126 0.374 contours of all samples in Table 1 are parameterized. Because
10 0.378 0.209
the contours of a beverage bottle are symmetrical, we study
...
61 0.765 0.485 half of the bottle. We use 16 key control points to quantify
62 0.454 0.528 the sample contours, as shown in Figure 5. All key points are
63 0.143 0.135 represented by coordinates, and some data are shown in
64 0.185 0.288 Table 12 (see Table S12 in the Supplementary Material for the
65 0.504 0.252 key point coordinates of all samples).
We take the data in Table 12 as input and the data in
Table 5 as output to train the ANN. The training results are
Table 7: The probability of the Kansei image cognition of the shown in Figures 6 and 7. Figures 6 and 7 show that the
design subjects. overall accuracies of the two ANNs are 92.374% and
87.731%, respectively, and that their errors are less than
Design subject Probability of the Kansei image cognition 0.125. Therefore, we obtain an ANN based on user cognition
User 0.378 and an ANN based on designer cognition. The two neural
Designer 0.367 networks are combined with the cognitive matching model
in Section 4.2, and this is used as the fitness function of the
product evolution system.
Table 8: The entropy, weight, and weighted cognitive evaluation of
the two design subjects. 4.5. Product Kansei Image Form Evolution System. We select
User Designer sample 38 from the sample set as the product prototype and
Entropy 0.963 0.973 form the parent sample with the representative Kansei image
Weigh 0.583 0.417 prototype. Combining the crossover probability and mu-
Weighted cognitive evaluation 1.848 1.284 tation probability determined in Section 4.1.3, we carry out
genetic operation and use the product Kansei image eval-
uation system obtained in the previous section to judge
We substitute the values in Table 10 into formula (14) whether the product form achieves cognitive matching. We
and coordinate index C and formula (14) into formula (21) use the MATLAB program to present the above operations
to obtain the cognitive matching model D based on the in the human-computer interaction interface, as shown in
design subjects. Figure 8.
12 Computational Intelligence and Neuroscience

Table 11: Kansei image prototype cluster.

1 3 4 5 6 8 10

11 12 13 15 17 18 19

22 26 27 38 39 40 42

44 48 51 54 55 61 62

65

Figure 4: The representative Kansei image prototype.

Computational Intelligence and Neuroscience 13

16 15
14

12 13
11
10

7
6
5
4
3
1 2
Figure 5: Key points of the beverage bottle contour.

After evolution, the product form with the highest evaluation coefficients of the evaluation system. The final
matching degree is selected. We save the coordinates of the product forms of all the Kansei image styles are shown in
key points, replace the product prototype to perform genetic Table 13.
operations again, and repeat until the product form yields a
cognitive matching degree D ≥ 0.75 between the two design
4.6. Verification of the Cognitive Matching Model. To verify
subjects.
the feasibility of the cognitive matching model proposed
In this process, the product evolution is led by the Kansei
in this paper, we combine all the product forms generated
image prototype, and the evolution process is similar to the
by the evolution system with the corresponding Kansei
communication between users and designers. Each iteration
images and invite 20 users and 20 designers to evaluate the
of the scheme is similar to the process of cognitive matching
Kansei image cognition. The results are shown in Table 14.
between the two design subjects. Because the product form
Combined with the results in Table 14, line charts of the
generated by the evolution system is used only as the form
predicted values and survey values of design subjects
reference for the designers in the design and in MATLAB,
under different Kansei images are drawn, as shown in
the linspace function is used for interpolation in the
Figures 14 ∼18.
transformation of the two-dimensional form into the three-
dimensional model, and the resulting contour of the three-
dimensional product model is not smooth. Therefore, we 4.7. Result Analysis. Figures 14∼18 show that, except for
derive the key points of the contours of all evolutionary “Comfortable,” the predicted results of the other four
products, use Rhinoceros software for modelling, and ar- Kansei images are basically consistent with the trend of
range all the schemes in the order of evolution to form the the survey results, and the evaluation results basically
evolution diagram of the Kansei image products, as shown in increase gradually. In addition, with the iteration of the
Figure 9. product scheme, the cognitive evaluations of the two
Because in the determination of the Kansei image design subjects basically cross each other, which is con-
prototype of the other four Kansei images the construction sistent with the matching process of the Kansei image
of the product Kansei image form evaluation system and the cognition of the two design subjects during the iteration of
product evolution process are the same as the above process, the product scheme. The cognitive evaluation of the
we repeat the process in 4.3∼4.5 to form the product evo- fourth-generation products is basically higher than that of
lution diagram of the other four Kansei images, as shown in the first three generations, which is consistent with the
Figures 10 ∼ 13. cognitive evaluation of the two design subjects being
Although the product prototype and Kansei image higher than that of the iterative process when the cog-
prototypes of different Kansei images are constructed nition achieves a high matching. Therefore, it is proven
under the same conditions, their evolution forms are that our cognitive matching model is feasible and can
different due to the different evolution probabilities and provide a reference for designers’ design activities.
 14

Table 12: Key point coordinates of some samples.

Point and
Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7 Sample 8 Sample 9
coordinate
x 0 0 0 0 0 0 0 0 0
1
y 0 0 0 0 0 0 0 0 0
x 23.89 26.37 37.15 31 13.47 24.17 34.57 3.81 34.34
2
y 0.1 0.1 0.1 0.1 -0.55 2.07 3.88 0.1 16.52
x 32.98 33.9 51.28 41.94 19.82 36.69 35.63 13.4 40.54
3
y 7.35 11.9 6.09 9.84 -1.56 27.12 11.55 -3.88 42.15
x 32.98 33.9 51.29 41.94 33.1 33.6 38.54 18.81 34.34
4
y 51.15 128.86 31.41 52.92 10.01 71.92 12.17 -5.92 75.2
x 32.98 29.94 54.22 41.94 34.16 30.32 39.16 31.33 31.04
5
y 105 146.3 49.08 96.31 21.03 102.71 22.49 3.47 102.95
x 32.98 29.94 57.88 41.94 34.16 29.99 34.35 31.33 31.04
6
y 148.76 162.31 76.35 147.64 83.29 125.28 37.53 85.73 115.18
x 20.22 30.6 55.21 41.94 34.16 32.26 33.2 30.39 41.86
7
y 172.44 174.51 107.56 180.98 129.1 155.17 82.46 97.19 145.05
X 14.63 26.76 36.17 41.94 31.52 35.1 33.78 28.8 40.45
8
y 181.33 215.48 151.92 206.11 164.14 179.08 135.73 106.01 150.93
x 14.63 24.29 28.98 35.32 25.57 31.85 38.01 28.8 40.57
9
y 218.36 231.83 173.9 218.45 190.18 203.07 212.87 180.09 162.51
x 14.63 20.12 27.14 43.44 18.93 19.04 16.58 31.46 37.06
10
y 249.53 245.51 191.56 218.55 217.71 222.65 231.16 194.38 193.36
x 17.36 20.52 28.98 43.44 15.29 10.76 14.2 13.63 25.16
11
y 249.53 249.28 211.71 237.07 232.34 231.47 236.45 231.95 220.31
x 17.36 20.52 32.63 43.44 13.17 10.76 14.2 13.63 25.16
12
y 250.62 255.32 227.29 254.71 247.06 236.66 243.58 239.54 224.44
x 13.78 22.6 32.63 43.44 15.29 14.11 17.55 16.32 26.34
13
y 252.59 256.22 239.85 269.22 247.16 236.76 243.68 239.54 224.54
x 13.78 22.6 35.79 37.66 15.29 14.11 17.55 16.32 26.34
14
y 267.29 262.06 250.12 274.73 263.15 253.16 259.29 255.41 257
x 13.78 29.23 28.99 21.72 10.12 10.76 13.41 10.47 21.85
15
y 269.06 265.63 259.57 274.63 266.35 256.24 263.34 259.72 262
x 0 0 0 0 0 0 0 0 0
16
y 269.16 265.73 259.67 274.73 266.45 256.34 263.44 260.7 264.76
Computational Intelligence and Neuroscience
Computational Intelligence and Neuroscience 15

Training: R=0.95498 Validation: R=0.99843

3.8
3.8

Output ~= 0.91∗Target + 0.26

Output ~= 2.8∗Target + -5.9

3.6
3.6

3.4
3.4

3.2 3.2

3 3

2.8 2.8
2.8 3 3.2 3.4 3.6 3.8 2.8 3 3.2 3.4 3.6 3.8
Target Target
Data Data
Fit Fit
Y=T Y=T

Test: R=0.89597 All: R=0.92374

3.8
3.8
3.6 Output ~= 0.9∗Target + 0.32
Output ~= 0.66∗Target + 1.2

3.6
3.4
3.4

3.2
3.2

3 3

2.8 2.8
2.8 3 3.2 3.4 3.6 3.8 2.8 3 3.2 3.4 3.6 3.8
Target Target
Data Data
Fit Fit
Y=T Y=T

Figure 6: Training results of the ANN based on user cognition.

According to Figure 17, there is a great difference be- used two-dimensional product contours, while the cognitive
tween the predicted value and the survey value of the design survey of the evolutionary product forms used a three-di-
subjects’ cognitive evaluation under “Comfortable.” The mensional model. Different product forms have an impact
possible reasons for this are as follows: (1) Compared with on the cognition of design subjects. Therefore, in subsequent
the other four Kansei images, “Comfortable” is more ab- studies, we will focus on the above issues in depth to improve
stract and is often expressed as tactile experience. It is the accuracy of the cognitive matching model.
difficult for design subjects to express its visual manifesta- In terms of innovation, existing studies, such as those in
tion, and most design subjects may still regard it as a tactile [19, 20, 28, 29], mainly analyse the cognitive differences from
experience. (2) The product form is the result of the synthesis a qualitative level and explore the final product form of
of multiple Kansei images. Because of the differences in the cognitive matching of different groups in product design,
degree of embodiment of different Kansei images, there are and they give less consideration to Kansei image traction.
primary and secondary distinctions between Kansei images. Compared with the existing studies, the advantages of
When we evaluate the product form of a single Kansei image, cognitive matching model of the design subjects constructed
the Kansei image may be affected by the primary Kansei in this paper are as follows:
image, which makes it difficult for the evolutionary form to
reflect the morphological characteristics of the Kansei image, (1) Based on the iteration of a single scheme, the
thus gradually reducing the cognitive evaluation of the cognitive change trend of users and designers in
Kansei image. In addition, the survey results in the process of scheme iteration is discussed
Figures 14∼18 are lower than the predicted results. This may quantitatively. This helps designers grasp the
occur because the previous cognitive survey of this study cognitive differences between the two cognitive
16 Computational Intelligence and Neuroscience

Training: R=0.88413 Validation: R=0.97048

4
3.4

Output ~= 0.62∗Target + 1.2

Output ~= 0.84∗Target + 0.4

3.5 3.2

3
3
2.8

2.5 2.6
2.5 3 3.5 4 2.6 2.8 3 3.2 3.4
Target Target
Data Data
Fit Fit
Y=T Y=T
Test: R=0.88865 All: R=0.87731
4
3.4 Output ~= 0.64∗Target + 1.1
Output ~= 1∗Target + 0.033

3.2 3.5

3
3
2.8

2.6 2.5
2.6 2.8 3 3.2 3.4 2.5 3 3.5 4
Target Target
Data Data
Fit Fit
Y=T Y=T
Figure 7: Training results of the ANN based on designer cognition.

subjects more accurately to improve the accep- 5. Discussion

tance of products.
In product design, the Kansei image cognition of users and
(2) To address the problem that the Kansei image style of designers is a factor that cannot be ignored. However, due to
the offspring product deviates from the style of the their different, complex backgrounds in terms of factors such
target Kansei image, this paper introduces the as education and society, there are differences between users’
concept of Kansei image traction and uses Kansei and designers’ cognitions; the designed products reflect the
image prototypes to guide the evolution of the off- designer’s cognition to a greater extent, resulting in low
spring product form to help ensure that the product recognition of products by users. The cognitive matching
scheme always has the characteristics of the target model of design subjects proposed in this paper simulates the
Kansei image. process of the cognitive matching of the two design subjects in
(3) The method proposed in this paper can effectively a quantitative way. By improving the cognitive matching
shorten the evaluation cycle of a large number of degree of design subjects in evolutionary products, we can
schemes generated by intelligent design, and it can reduce the cognitive differences in the process of product
help designers quickly and accurately select alter- design. To some extent, the matching model helps designers
native schemes that fit the Kansei image cognition of understand the cognitive trends of the two design subjects and
the two design subjects, thus shortening the product comprehensively consider their own cognition and users’
development cycle to a certain extent. needs so that a strong matching of the Kansei image cognition
Computational Intelligence and Neuroscience 17

Figure 8: Human-computer interaction interface of the product evolution system.

Product Prototype Iteration 1 Iteration 2 Iteration 3 Iteration 4 Kansei Image Archetype

Figure 9: The product evolution process of “Exquisite.”

Product Prototype Iteration 1 Iteration 2 Iteration 3 Iteration 4 Kansei Image Archetype

Figure 10: The product evolution process of “Simplism.”

Product Prototype Iteration 1 Iteration 2 Iteration 3 Iteration 4 Kansei Image Archetype

Figure 11: The product evolution process of “Streamlined.”

of design subjects and the efficient development of products Human cognition of things is a complex process that is
can be achieved. This provides guidance and new research affected by many factors such as regional culture, educa-
ideas for intelligent product design that balances the cognition tional background, and the surrounding environment. In
of multiple design subjects. this paper, a questionnaire survey is used to obtain the
18 Computational Intelligence and Neuroscience

Product Prototype Iteration 1 Iteration 2 Iteration 3 Iteration 4 Kansei Image Archetype

Figure 12: The product evolution process of “Comfortable.”

Product Prototype Iteration 1 Iteration 2 Iteration 3 Iteration 4 Kansei Image Archetype

Figure 13: The product evolution process of “Novel.”

Table 13: The evolution forms of the final products from different images.

Exquisite Simplism Streamlined Comfortable Novel

Table 14: Comparison of the survey evaluation and predicted evaluation results.
Kansei image Design subject 1 2 3 4
User (survey) 2.9 3 3.1 3.55
User (prediction) 3.659 3.951 4.109 4.215
Exquisite
Designer (survey) 3.1 3.15 3.05 3.25
Designer (prediction) 3.015 3.267 3.653 4.051
User (survey) 2.4 2.7 3.15 3.65
User (prediction) 4.441 3.489 4.167 4.915
Simplism
Designer (survey) 2.75 2.55 3.2 3.25
Designer (prediction) 3.295 3.956 3.996 4.297
User (survey) 2.75 2.65 3.5 3.6
User (prediction) 3.951 4.426 4.686 4.712
Streamlined
Designer (survey) 3.05 2.8 3.4 3.75
Designer (prediction) 3.27 3.568 4.009 4.57
User (survey) 3 3.1 2.9 3.2
User (prediction) 3.578 4.239 4.391 4.504
Comfortable
Designer (survey) 3.6 3.1 3 2.95
Designer (prediction) 3.616 3.752 4.164 4.68
User (survey) 2.8 3.222 3.3 3.394
User (prediction) 2.9 3.807 3.45 4.269
Novel
Designer (survey) 3.45 4.591 3.9 4.083
Designer (prediction) 4.05 4.863 4.4 4.661
Computational Intelligence and Neuroscience 19

Evaluation value
3

1
1 2 3 4
Iterations

User (Investigative) Designer (Investigative)

User (predictive) Designer (predictive)
Figure 14: Comparison of the cognitive evaluation under “Exquisite.”

4
Evaluation value

1
1 2 3 4
Iterations
User (investigative) Designer (investigative)
User (predictive) Designer (predictive)
Figure 15: Comparison of the cognitive evaluation under “Simplism.”

cognition of design subjects, and this process has strong An accurate product model can more accurately reflect
subjectivity. In future research, we will apply more objective the Kansei image cognition of design subjects. In this paper,
methods such as EEG experiments to obtain the Kansei the linspace function can roughly reflect the three-dimen-
image cognition of design subjects. sional shape of the product, but the contour fluency of the
The Kansei image prototype is the concrete form of the model is not too high. Subsequent research will further
Kansei image, and morphological fusion needs to consider the explore the interpolation function of the model construction
cognitive commonality and the topological structure between in the three-dimensional modelling software in order to
the two characteristic lines. To facilitate the fusion between redesign the product Kansei image form evolution system to
forms, the beverage bottle with the highest Kansei image improve the accuracy of the three-dimensional model.
evaluation is regarded as the Kansei image prototype. Future In the process of product development, all participants,
research will focus on the fusion algorithm of different types of such as users, designers, engineers, design decision-makers,
forms, thus simplifying the process of morphological fusion. and marketers, are the design subjects of the product. To
20 Computational Intelligence and Neuroscience

Evaluation value
3

1
1 2 3 4
Iterations

User (investigative) Designer (investigative)

User (predictive) Designer (predictive)
Figure 16: Comparison of the cognitive evaluation under “Streamlined.”

4
Evaluation value

1
1 2 3 4
Iterations

User (investigative) Designer (investigative)

User (predictive) Designer (predictive)
Figure 17: Comparison of the cognitive evaluation under “Comfortable.”

facilitate this research, the design subjects are only the users with a small sample size, using a small sample size in training
and designers in the paper, and other participants such as to improve the recognition performance of the evaluation
engineers are not considered. Next, we will introduce other model.
design participants to explore the impact of different par- Finally, colour, material, and surface technology (CMF)
ticipants’ interventions on product form. are popular research topics at this stage, and the integration
The evaluation system of this study is based on an ANN, of various elements can reflect the cognition of design
which needs a large number of samples to improve the subjects. However, to improve the accuracy of the cognitive
prediction accuracy. In this study, the sample size used in matching model, this paper explores only the cognitive
neural network training is relatively small, which may re- matching process of the product form. Combining CMF to
duce the accuracy of the evaluation system. Future research further explore cognitive matching from the overall per-
will focus on exploring the intelligent evaluation algorithm spective will be the focus of future research.
Computational Intelligence and Neuroscience 21

Evaluation value
3

1
1 2 3 4
Iterations

User (investigative) Designer (investigative)

User (predictive) Designer (predictive)
Figure 18: Comparison of the cognitive evaluation under “Novel.”

6. Conclusions Data Availability

This paper presents a method of cognitive matching of The data used to support the findings of this study are in-
design subjects in product form evolutionary design. First, cluded within the supplementary information files.
we explore the inherent trends of the Kansei image cognition
of the two design subjects by using the relevant methods of Conflicts of Interest
Kansei engineering, and we construct a cognitive matching
The authors declare that there are no conflicts of interest
model of the design subjects. Second, according to the
regarding the publication of this paper.
obtained product form that represents the Kansei image, the
Kansei image prototype cluster is constructed, and the Acknowledgments
representative Kansei image prototype is selected as external
traction to lead the evolution of the product form. Then, we This research was funded by the National Natural Science
combine an ANN with the cognitive matching model of the Foundation of China (Grant no. 51705226), Natural Science
design subjects to construct the product Kansei image Foundation of Gansu Province (Grant no. 20JR10RA168), and
evaluation system; this is used as the evaluation mechanism Gansu Provincial Department of Education: Outstanding
of the evolution system. Finally, a product Kansei image Postgraduate “Innovation Star” Project (2021CXZX-524).
form evolution system is established based on a GA and
verified by comparative experiments. The results show that Supplementary Materials
the model can effectively realize the Kansei image cognition
of the two design subjects. Table S1: all representative samples. Table S3: survey results
The proposed method largely simulates the process of of the Kansei image cognition of all users. Table S4: survey
the interaction between designers and users in product results of the Kansei image cognition of all designers. Table
design, helps designers to more accurately understand the S5: all evaluation values of “Exquisite.” Table S6: all nor-
cognitive trends of the two design subjects, and provides a malization results. Table S12: key point coordinates of all
quantitative method for the cognitive matching of the design samples. (Supplementary Materials)
subjects, which has guiding significance for the continuation
of the evolutionary product Kansei image style in the process References
of product form evolution. Although we take beverage
[1] Y. Xiong, Y. Li, and G. L. Ma, “Productform design based on
bottles as an example to study, this model is applicable to the
the sketch cognitive analysis,” Computer Integrated
form evolution of other products. Simulating the cognitive Manufacturing Systems, vol. 16, no. 7, pp. 1354–1362, 2010.
matching process of the two design subjects in the iterative [2] J. N. Su, Y. J. Kang, and S. T. Zhang, “Innovative design
process of the product scheme provides a research basis for method of product image styling oriented to cognitive sub-
intelligent product design that balances the cognition of ject,” Modern Manufacturing Engineering, no. 6, pp. 108–113,
multiple design subjects. 2018.
22 Computational Intelligence and Neuroscience

[3] J. Vieira, J. M. A. Osório, S. Mouta et al., “Kansei engineering [22] D. Jeremiah, “ Still, and V. J. Dark, “Cognitively describing
as a tool for the design of in-vehicle rubber keypads,” Applied and designing affordances,” Design Studies, vol. 34, no. 3,
Ergonomics, vol. 61, pp. 1–11, 2017. pp. 285–301, 2013.
[4] K. Qiu, J. Su, X. Zhang, and W. Yang, “Evaluation and balance [23] T. Kujala and P. Saariluoma, “Cognitive mimetics for de-
of cognitive friction: evaluation of product target image form signing intelligent technologies,” Advances In Human-Com-
combining entropy and game theory,” Symmetry, vol. 12, puter Interaction, vol. 2018, Article ID 9215863, 9 pages, 2018.
no. 9, p. 1398, 2020. [24] N. Zane, S. Sue, J. Chang et al., “Beyond ethnic match: effects
[5] T. T. Hu, J. H. Zhao, and D. H. Zhao, “6 Circulation design,” of client-therapist cognitive match in problem perception,
Car Park Designers’ Handbook, vol. 36, no. 24, pp. 33–36, coping orientation, and therapy goals on treatment out-
2015. comes,” Journal of Community Psychology, vol. 33, no. 5,
[6] S. W. Hsiao and H. C. Tsai, “Applying a hybrid approach pp. 569–585, 2005.
based on fuzzy neural network and genetic algorithm to [25] C. Lucchiari and G. Pravettoni, “Cognitive balanced model: a
product form design,” International Journal of Industrial conceptual scheme of diagnostic decision making,” Journal of
Ergonomics, vol. 35, no. 5, pp. 411–428, 2004. Evaluation in Clinical Practice, vol. 18, no. 1, pp. 82–88, 2012.
[7] W. J. Yang, J. N. Su, and S. T. Zhang, “Intelligent design of [26] J. Roth, M. C. Steffens, and V. L. Vignoles, “Group mem-
product forms based on design cognitive dynamics and a bership, group change, and intergroup Attitudes: a recate-
cobweb structure,” Computational Intelligence and Neuro- gorization model based on cognitive consistency principles,”
science, vol. 2021, Article ID 6654717, 17 pages, 2021. Frontiers in Psychology, vol. 9, p. 479, 2018.
[8] J. Shi, J. N. Su, and X. Li, “Coupling optimization design [27] A. G. O. Hassan and K. Ralf, “Accounting and politicians: a
research of product image prototype form,” Packaging En- theory of accounting information usefulness,” Journal of
gineering, vol. 40, no. 8, pp. 47–53+129, 2019. Public Budgeting, Accounting and Financial Management,
[9] H.-B. Yan, V.-N. Huynh, T. Murai, and Y. Nakamori, “Kansei vol. 31, no. 4, pp. 496–517, 2019.
evaluation based on prioritized multi-attribute fuzzy target- [28] S. Luo and S. S. Zhu, “Users’ and designers’ product form
oriented decision analysis,” Information Sciences, vol. 178, perceptual image,” Chinese Journal of Mechanical Engineer-
no. 21, pp. 4080–4093, 2008. ing, vol. 41, no. 10, pp. 28–34, 2005.
[10] M. Kohler, B. Falk, and R. Schmitt, “Applying eye-tracking in [29] J. N. Su, Y. H. Shen, and W. J. Yang, “Product form innovation
Kansei engineering method for design evaluations in product design based on cognitive thinking and cobweb structure,”
development,” International Journal of Affective Engineering, Packaging Engineering, vol. 42, no. 2, pp. 14–21, 2021.
vol. 14, no. 3, pp. 241–251, 2015. [30] M. Aslam, A. Fahmi, F. A. A. Almahdi, and N. Yaqoob,
[11] J. Yan and B. Oren, “Creative patterns and stimulation in “Extension of TOPSIS method for group decision-making
conceptual design,” Analysis and Manufacturing, vol. 24, under triangular linguistic neutrosophic cubic sets,” Soft
no. 2, pp. 191–209, 2010. Computing, vol. 25, no. 5, pp. 3359–3376, 2021.
[12] L. Xue, X. Yi, and Y. Zhang, “Research on optimized product [31] M. Zaree, S. Javadi, and A. Neshat, “Potential detection of
image design integrated decision system based on Kansei water resources in karst formations using APLIS model and
engineering,” Applied Sciences, vol. 10, no. 4, p. 1198, 2020. modification with AHP and TOPSIS,” Journal of Earth System
[13] S. T. Zhang, P. F. Su, and S. F. Liu, “Fusion of cognitive Science, vol. 128, no. 4, pp. 1–12, 2019.
information: evaluation and evolution method of product [32] A. K. Yadav and R. Srivastava, “Selection of teak sawdust
image form,” Computational Intelligence and Neuroscience, polypropylene composite’s composition for outdoor appli-
vol. 2021, Article ID 5524093, 19 pages, 2021. cations using TOPSIS analysis,” S�adhan�a, vol. 45, no. 1, p. 231,
[14] B. Q. Xu, Y. N. Zhu, and X. J. Liu, “Comparison of product 2020.
image gene network between designers and users,” Packaging [33] M. Qiyas, S. Abdullah, Y. D. Al-Otaibi, and M. Aslam,
Engineering, vol. 41, no. 8, pp. 98–105, 2020. “Generalized interval-valued picture fuzzy linguistic induced
[15] D. Zhong, Y. A. Zhu, and L. Q. Wang, “A cognition hybrid operator and TOPSIS method for linguistic group
knowledge representation model based on multidimensional decision-making,” Soft Computing, vol. 25, no. 7,
heterogeneous data,” Complexity, vol. 2020, Article ID pp. 5037–5054, 2021.
8812459, 17 pages, 2020. [34] Y. Wang, J.-R. Chardonnet, and F. Merienne, “Enhanced
[16] D. D. Ye and S. G. Li, “Hierarchy of needs and design strategy cognitive workload evaluation in 3D immersive environments
in interaction design,” Packaging Engineering, vol. 34, no. 8, with TOPSIS model,” International Journal of Human-
pp. 75–78, 2013. Computer Studies, vol. 147, Article ID 102572, 2021.
[17] Y. L. Lee and L. C. Chen, “Why social networking?” Social [35] H. Farajpanah, M. Lotfirad, A. Adib et al., “Ranking of hybrid
Software Engineering, vol. 5, no. 1, pp. 16–27, 2016. wavelet-AI models by TOPSIS method for estimation of daily
[18] Y. Cao and Y. Liu, “A brief analysis on the mental model flow discharge,” Water Supply, vol. 20, no. 8, pp. 3156–3171,
matching between designers and users,” Decoration, no. 6, 2020.
pp. 98-99, 2011. [36] K. Naeem, M. Riaz, and X. Peng, “Pythagorean m-polar fuzzy
[19] Y. Yao and Y. Q. Huang, “Research on the mapping model topology with TOPSIS approach in exploring most effectual
between users’ cognitive concept and product shape infor- method for curing from COVID-19,” International Journal of
mation,” Applied Mechanics and Materials, vol. 556-562, Biomathematics, vol. 13, no. 8, Article ID 2050075, 2020.
pp. 4051–4054, 2014. [37] J.-Y. Dong and S.-P. Wan, “Virtual enterprise partner se-
[20] C. X. He, “Analysis of cognitive friction in product design and lection integrating LINMAP and TOPSIS,” Journal of the
use process,” Packaging Engineering, vol. 31, no. 8, pp. 58–60, Operational Research Society, vol. 67, no. 10, pp. 1288–1308,
2010. 2016.
[21] Z. Y. Tan, L. Liu, and Y. C. Jiang, “Behavior style based on [38] S.-P. Wan, F. Wang, L.-L. Lin, and J.-Y. Dong, “Some new
cognitive conflict during interaction design,” Packaging En- generalized aggregation operators for triangular intuitionistic
gineering, vol. 40, no. 6, pp. 165–172, 2019. fuzzy numbers and application to multi-attribute group
Computational Intelligence and Neuroscience 23

decision making,” Computers & Industrial Engineering,

vol. 93, pp. 286–301, 2016.
[39] S.-p. Wan, G.-l. Xu, and J.-y. Dong, “A novel method for
group decision making with interval-valued Atanassov
intuitionistic fuzzy preference relations,” Information Sci-
ences, vol. 372, pp. 53–71, 2016.
[40] S.-P. Wan, G.-l. Xu, F. Wang, and J.-y. Dong, “A new method
for Atanassov’s interval-valued intuitionistic fuzzy MAGDM
with incomplete attribute weight information,” Information
Sciences, vol. 316, pp. 329–347, 2015.
[41] H. Lin, T. H. Pan, and S. Chen, “Comprehensive evaluation of
urban air quality using the relative entropy theory and im-
proved TOPSIS method,” Air Quality, Atmosphere & Health,
vol. 14, no. 2, pp. 1–8, 2021.
[42] M. Saeli, R. Micale, M. P. Seabra, J. A. Labrincha, and G. La
Scalia, “Selection of novel geopolymeric mortars for sus-
tainable construction applications using fuzzy topsis ap-
proach,” Sustainability, vol. 12, no. 15, p. 5987, 2020.
[43] J. Zhou, T. F. Xiahou, and Y. Liu, “Multi-objective optimi-
zation-based TOPSIS method for sustainable product design
under epistemic uncertainty,” Applied Soft Computing Jour-
nal, vol. 98, no. 8, Article ID 106850, 2021.
[44] W. J. Wang, S. H. Yu, and S. X. Wang, “Multilevel com-
prehensive evaluation and decision making of ergonomics,”
Discrete Dynamics in Nature and Society, vol. 2015, Article ID
689203, 9 pages, 2015.
[45] V. Tiwari, P. K. Jain, and P. Tandon, “An integrated Shannon
entropy and TOPSIS for product design concept evaluation
based on bijective soft set,” Journal of Intelligent
Manufacturing, vol. 30, no. 4, pp. 1645–1658, 2019.
[46] J. M. Xie, Q. Qin, and M. T. Jiang, “Multiobjective decision-
making for technical characteristics selection in a house of
quality,” Mathematical Problems in Engineering, vol. 2020,
Article ID 9243142, 12 pages, 2020.
[47] J. N. Su, Y. H. Chen, and N. Jing, “Coupling characteristics
study in product image modeling design,” Journal of Machine
Design, vol. 34, no. 1, pp. 105–109, 2017.
[48] J. N. Su, X. X. Zhang, and N. Jing, “Research on the entropy
evaluation of product styling image under the cognitive
difference,” Journal of Machine Design, vol. 33, no. 3,
pp. 105–108, 2016.
[49] S.-P. Wan, Q.-Y. Wang, and J.-Y. Dong, “The extended
VIKOR method for multi-attribute group decision making
with triangular intuitionistic fuzzy numbers,” Knowledge-
Based Systems, vol. 52, pp. 65–77, 2013.
[50] S.-P. Wan, Z. Jin, and J.-Y. Dong, “Pythagorean fuzzy
mathematical programming method for multi-attribute
group decision making with Pythagorean fuzzy truth de-
grees,” Knowledge and Information Systems, vol. 55, no. 2,
pp. 437–466, 2018.
[51] S.-p. Wan, W.-c. Zou, L.-g. Zhong, and J.-y. Dong, “Some new
information measures for hesitant fuzzy PROMETHEE
method and application to green supplier selection,” Soft
Computing, vol. 24, no. 12, pp. 9179–9203, 2020.
[52] G.-l. Xu, S.-P. Wan, and J.-Y. Dong, “An entropy-based
method for probabilistic linguistic group decision making and
its application of selecting car sharing platforms,” Informa-
tica, vol. 31, no. 3, pp. 621–658, 2020.
[53] S. T. Zhang, P. F. Su, and W. J. Yang, “Integrated evaluation
method of product aesthetic based on entropy theory,”
Packaging Engineering, vol. 42, no. 8, pp. 79–87, 2021.