Perception modeling based on vision and touch: “authenticity” of leather defined through observational experience

Abstract

Nowadays, numerous products use artificial leather as it is a cost-effective alternative to genuine leather. However, products made from artificial leather may leave impressions on consumers that are dissimilar to those left by products made of genuine leather. In other words, products that use artificial leather but are perceived as genuine leather are more attractive to consumers. Therefore, in this study, we aimed to understand and quantify the factors that affect the mechanism via which consumers perceive a leather product to be made of genuine leather. We conducted several experiments to evaluate the hypothesis regarding human perception. Measurement experiments were performed to obtain the visual and physical properties of such impressions. We estimated the representative impressions formed by people during their interaction with leather samples through subjective experiments and derived models of these impressions in terms of the measured properties. Subjective evaluation experiments were performed under visual, tactile, and visual–tactile conditions. Finally, we quantified leather “authenticity” using these representative impressions. Participants, who are general consumers, were divided into two groups according to their familiarity with leather. The “authenticity” perception model of the group familiar with leather was constructed under visual and visual–tactile conditions, whereas the model of the group unfamiliar with leather was constructed under visual–tactile conditions, suggesting the influence of a cross-modal phenomenon. The results of this study can be applied to develop attractive artificial leather, which is expected to contribute to the protection of animal rights while promoting the sale of artificial leather products.

Keywords

Quality fabrication measurement materials human sensory and comfort issues management of systems product and systems engineering product design consumer behavior/demand social science product meaning and consumer perception fashion identity branding

We form various impressions of objects by observing their material properties in our everyday life. In addition, we can sense various qualities of materials known as shitsukan.¹ This term encompasses topics relating to material perception and object recognition.²,³ Using shitsukan information, we can determine if an object is fragile and how carefully it should be handled. In contrast, if we discern that a road is slippery by looking at its surface features, we adapt our behavior accordingly, that is, by walking slowly and carefully.⁴ The motivation of a consumer when purchasing a particular product is similarly influenced by their impression or shitsukan (for example, “favorable/unfavorable”) of that product.⁵,⁶ Therefore, numerous studies on understanding human perception mechanisms that elucidate shitsukan have been conducted.⁷,⁸ In addition, the properties of products that maximize shitsukan must be quantified and controlled. Thus, methods for identifying and quantifying these properties for various materials have been proposed.^10–12

In the present study, various leather samples are investigated to focus on the human perception of leather. Product designers often prefer using genuine leather owing to its superior quality. However, artificial leather is widely used owing to the high price of genuine leather and the environmental issues and animal-rights concerns associated with the leather industry.¹³,¹⁴ Artificial leather is manufactured to resemble genuine leather. However, products made from artificial leather may leave impressions on consumers that are different from those left by products made of or comprising genuine leather.

We therefore believe that if we elucidate the psychological mechanism used by humans to perceive leather and identify the physical properties associated with this mechanism, it would be possible to manufacture artificial leather that induces shitsukan, that is, “authenticity (artificial leather that is perceived as genuine leather).” Therefore, the aim of this study was to identify the impressions and properties that consumers perceive from leather and to ultimately quantify the “authenticity.” Quantifying the “authenticity” can contribute to the development of artificial leather that is highly attractive to consumers. In addition, in this study, considering previous studies that indicated that human perception is influenced by their experience with the object, we divided the participants into two groups according to their familiarity with leather.¹⁵ In our previous studies, we examined the quantification of “authenticity” as perceived by the human eye and other aspects of shitsukan.¹⁶,¹⁷ However, consumers evaluate leather products based on not only visual but also tactile information. Thus, in the present study, we examined the quantification of “authenticity” by conducting experiments under visual, tactile, and visual–tactile conditions.

In terms of related studies, a European Union project, “Measurement of Naturalness,” studied the relationships between the physical properties of natural and synthetic materials and the sensory processes that lead to the perceptual judgment of “naturalness.”^18–20 However, in this project, the aim was to distinguish whether it can be perceived as natural or not by using samples made from natural, artificial, and mixed materials. On the contrary, in the present study, we first hypothesized a hierarchy of subjective responses of leather “authenticity.” Furthermore, some studies constructed a hierarchical model using impression words prepared in advance.^21–25 The low correlation between each level was attributed to the uncertainty of whether the prepared impression words are appropriate for the purpose of evaluation. In contrast, we collected impression words from the participants and then extracted highly suitable impression words for leather by several subjective evaluation experiments. Therefore, a high correlation is expected between each level. In addition, several studies were only concerned with the correlation between the words expressing the impressions.^23–25 In contrast, we also examined the correlation between the words and the leather properties. Identification of the relationship between the physical properties of leather and human subjectivity can be useful in industry. Kawabata et al.^26–29 published many studies on the perception of textiles. They mainly explored the tactile sensations directly related to physical properties, such as stiffness and smoothness. In contrast, we explore shitsukan, which consumers perceive via integration of tactile sensations. Moreover, the tactile sensations we deal with in this study are not specific in nature, but are estimated through subjective experiments.

Perception model

The process of receiving a sensory stimulus to perceiving it does not occur in a single step. In this regard, Schiffman³⁰ and Hui and Sherratt³¹ described a three-step mechanism, involving stimulus, sensory, and perception. The sensory step is related to the perception of qualities or attributes related to the physical environment produced by the stimulus. Perception refers to the phenomena created by psychological processes such as the integration of sensations derived from stimuli.

Previous brain science studies have reported that when stimuli detected by the retina and mechanoreceptors reach the brain, the simple information is processed in the early stages by the primary visual cortex (V1) and primary somatosensory cortex (S1) and complex information is processed in the later stages by, among others, the inferior temporal cortex (IT).^32–35 Furthermore, in studies dealing with multiple sensory information, it was reported that each sensor does not function independently but affects other sensors.³⁶

Based on the above-mentioned studies, we hypothesize the hierarchy of subjective responses shown in Figure 1. In the figure, “Leather properties” (e.g., color, and roughness) correspond to the stimuli detected by humans and were obtained via measurement experiments. Sensory derived data informing “Impression” (e.g., softness, and silkiness), “authenticity,” and perception of shitsukan were obtained during several subjective evaluation experiments. Here, “Impression” encompasses the representative impressions we form of leather. We apply this hierarchy to confirm the correlation between leather properties and subjective responses.

Figure 1.
The subjective response hierarchy of perception.

Measurement experiments

To measure the visual and physical properties associated with perception, we used gonio-spectrophotometer systems developed inhouse and several commercial measurement devices. Figure 2 shows the enlarged images of the surfaces of 10 leather test samples used in this study. No. 1–4 are images of cowhide leather and no. 5–10 are images of artificial leather. All the samples were black, which is also the most common color of leather products, and each sample size was approximately 300 mm × 210 mm. The RGB images were captured under D65 diffuse illumination.

Figure 2.
Surface images of genuine and artificial leather test samples. The insets reflect enlarged images of the sample surface.

Measurement of visual properties

We conducted two types of experiments: color property measurement and surface characterization.¹⁶,¹⁷ Figure 3 shows a schematic of each experimental system.

Figure 3.
Configuration of each experimental system: (a) color measurement system; (b) surface characterization system. DSLR: digital single-lens reflex.

The color measurement system consisted of a spectral camera and a light source (xenon). The spectral camera captured images in 31 bands with a 10-bit depth. The size and resolution of the captured images were 600 × 600 pixels and approximately 1000 dpi, respectively. The spectral images were converted into CIE Lab* (where L* is the lightness value, and a* and b* represent the color channels) format and normalized by using spectral images of a standard white target.³⁷ The measurement angle was 45°, and the illumination angles were –15°, 0°, 20°, 30°, and 45° from the normal direction (corresponding geometries are (light-source angle/detector angle) –15°/45°, 0°/45°, 20°/45°, 30°/45°, and 45°/45°). We set these angles with reference to commonly used color measurement angles. In addition, the visual properties, kurtosis, and distortion of the L* were calculated for each geometry.

The surface characterization system comprised a digital single-lens reflex (DSLR) camera (PENTAX K-3 II, RICOH IMAGING COMPANY, LTD, Ota-ku, Tokyo, Japan) and a lighting device. The raw images had a color depth of 14 bits. The raw images were then normalized by using raw images of a standard white target. The size of the captured image was trimmed to 3000 × 3000 pixels. The image resolution was approximately 1000 dpi. The measurement angle was 0°, and the illumination angles were 15°, 25°, 45°, and 60° from the normal direction (corresponding geometries were 15°/0°, 25°/0°, 45°/0°, and 60°/0°, respectively).

Using the image obtained from each angle, the visual properties, that is, surface characteristics, spatial frequency, and grain size of the leather surfaces, were calculated.

To obtain the surface characteristic values, we performed a two-dimensional (2D) Fourier transform on the ΔL* image, which was obtained by subtracting the average value from the L* image. Since the average value of the ΔL* image was 0, we could compare the surface characteristics using only the brightness contrast. For conversion into the one-dimensional (1D) characteristics of spatial frequency, the cyclic average values for each spatial frequency (cycles/mm) were calculated (Figure 4 shows this step). In the graph, the vertical axis is the amplitude and the horizontal axis represents spatial frequency.

Figure 4.
Procedure used to calculate spatial frequency characteristics. 2D: two-dimensional.

The obtained spatial frequency characteristics were weighted by the human contrast sensitivity characteristics, referred to as the contrast sensitivity function (CSF). Numerous CSF models have been proposed; however, in this study, we used the basic model proposed by Dooley and Shaw.³⁸ The surface of leather has various frequency bands because of its irregularity and varying grain size. Integral values of the spatial frequencies within the ranges of 0–0.1, 0.1–1.0, and 1.0–4.0 cycles/mm were defined as surface characteristic values; these values corresponded to surface irregularity, large grains, and small grains, respectively. The ranges were determined by measuring the grain sizes of the samples used.

The procedure for calculating the grain size was as follows. Firstly, the L* image was binarized to distinguish the grains from their background. In this study, the threshold value was determined by the P-tile method (P = 25%). We then performed contour extraction on the binarized image and extracted the grains. Finally, the number of pixels constituting each grain was counted, and we defined the grain size as the average number of pixels.

Measurement of physical properties

To obtain the physical properties, we used the Kawabata evaluation system (KES, Kato Tech. Co., Ltd, Kyoto-shi, Kyoto, Japan). The KES is widely used in textile studies to quantify the properties of textiles,³⁹,⁴⁰ and is also used in leather studies.⁴¹,⁴² Tactile sensation is stimulated mainly by a combination of the friction, roughness, elasticity, and the thermal properties of an object.⁴³ We measured each physical property of the samples, accordingly. The measurements were performed in an environment with a temperature of 22°C and a humidity of 65%.

Friction and roughness were measured with a surface tester (KES-FB4-A). This device calculates the MIU (average dynamic friction coefficient), MMD (variation of average dynamic friction coefficient), and SMD (average deviation of surface roughness) by the round-trip scanning of the sample surface with a sensor. Furthermore, the sensor obtained the forward, return, and round-trip values. Subsequently, we calculated the maximum static friction coefficient from the measurement data. Each sample was measured three times.

The thermal properties were measured using a thermal measurement instrument (KES-F7). This device determines the q_max value (peak heat flux). A large q_max value indicates intense instantaneous coldness. We measured each sample three times.

Elasticity was measured with a compression tester (KES-FB3-A). This device calculates the LC (compressional linearity), WC (compressional energy), and RC (compressional recoverability) by pressurizing and depressurizing the sample surface with a sensor. We measured each sample six times.

Table 1 lists the 59 visual and physical properties obtained in this study.

Table 1.
List of visual and physical properties obtained under different methods: (a) visual properties; (b) physical properties

(a)Method Properties

Color property measurement(5 geometries) - Lab* value- Skewness and kurtosis of L*

Surface characterization(4 geometries) - Average grain size- Integrated value of spatial frequency band (irregularity, large grain, and small grain components) - Integrated value of the entire spatial frequency range

(b)Device Properties

KES-FB4-A Surface Tester Values obtained for the forward, return, and round trip1. MIU (average dynamic friction coefficient)2. MMD (variation of average dynamic friction coefficient) 3. SMD (average deviation of surface roughness) - Maximum static friction coefficient

KES-F7 Thermo Labo - q_max value (peak heat flux)

KES-FB3-A Compression Tester - LC (compression linearity) - WC (compression energy) - RC (compression recoverability)

Figure 5 shows some of the raw data of the visual and physical properties obtained during these measurement experiments. These property data are used in the multiple regression analyses in the fifth section.

Figure 5.
Visual properties obtained during the measurements. These properties are used in multiple regression analyses in the fifth section: (a) visual properties; (b) physical properties. MMD: variation of average dynamic friction coefficient; SMD: average deviation of surface roughness.

Subjective evaluation experiments

Impression factors

Factor analysis was used to estimate the representative impressions of leather. Adjectives used for the factor analysis were extracted using the following subjective evaluation experiments.¹⁶,¹⁷

Word extraction experiment

Under visual conditions, 10 male participants, 20–40 years of age, evaluated the samples and listed as many words as they could think of to describe their impressions. Leather products were not targeted to a specific generational market; therefore, we selected participants representative of a wide age group. They were our colleagues who were typical leather consumers and the average visual acuity, including correction, was approximately 1.0. In contrast, under tactile and visual–tactile conditions, 11 people (the 10 participants under visual conditions and one additional male) participated. They were not aware of the purpose of this study. In this experiment and the subsequent experiments, the participants evaluated the samples under the conditions presented in Figure 6. Under visual and visual–tactile conditions, we used a standard light source (Spectra Light QC; X-rite, Grand Rapids, MI, USA). The standard illuminant D65 was used and the illuminance was set to 1270 lx. The observation distance was 300 mm from the sample and the participants maintained the observation distance by resting their foreheads against a prop fixed in the appropriate position. The samples were displayed in a curved state (with a curvature radius of approximately 190 mm). Therefore, the participants could observe the samples under various angle conditions from a fixed position. Under tactile conditions, the leather samples were hidden from the participants’ view, while allowing stroking of the samples with palms and fingers. As a result, 92, 94, and 117 words were extracted under visual, tactile, and visual–tactile conditions, respectively. In a previous study, the test conducted on 10 participants had a word duplication rate of 84%, with most leather-related words extracted.²⁴

Figure 6.
Subjective experimental conditions: (a) visual conditions; (b) tactile conditions; (c) visual–tactile conditions.

Suitable word experiment

In this experiment, we divided the participants into two groups: those who were familiar with leather and those who were unfamiliar with leather. The participants with the following qualities belonged to the familiar group: (a) owned leather products (that is, those who have had several opportunities to observe leather products); (b) interested in leather (that is, those who have keenly observed leather products). The participants evaluated the suitability of words, with regard to their impressions of leather, using a scale from 1 to 7. Under visual conditions, 20 participants aged 20–60 years participated in the experiment; 10 of them had participated in the previous test (eight familiar and 12 unfamiliar with leather). Under tactile and visual–tactile conditions, 25 participants aged 20–60 years participated, including 10 from the previous test (10 familiar and 15 unfamiliar with leather). The participants determined the appropriateness of words with each sample under visual, tactile, and visual–tactile conditions to evaluate 92, 94, and 117 ratings, respectively. We selected the suitable words under the following conditions: (a) the average score was more than 5; (b) the standard deviation score of the word was lower than the average standard deviation scores +1σ of the aforementioned high average words. Figure 7 shows the results of the experiment with a familiar group under visual conditions. Of the 92 words that were plotted, only those in the red area were selected.

Figure 7.
Results of the experiment to select suitable words conducted with the familiar group under visual conditions. (Color online only.)

Consequently, under visual conditions, we extracted 24 and 28 words by the familiar and unfamiliar group, respectively. Furthermore, under tactile and visual–tactile conditions, we extracted 13 and 31 words, and 28 and 39 words, respectively. We inferred that the group unfamiliar with leather often evaluated the words as appropriate because they could not identify the difference between similar descriptors for leather.

Word distance experiment

The participants evaluated the similarity of all word pairs. The participants determined whether the distance between each word pair was a “short-distance” or “long-distance.” Subsequently, the ratio of the participants that answered “long-distance” was defined as the distance between each word pair. In other words, if the ratio was “1,” the two words are independent. If the ratio was “0,” the two words have the same meaning. The test was conducted using an Excel sheet that randomly listed the number of suitable words C₂ word pairs.

We then plotted the words using multidimensional scaling (MDS) for the distance between each word pair. Next, hierarchical cluster analysis was performed on the dimensionally plotted words. In this study, we combined words that were closer together than the average inter-cluster distance. As a result, under visual conditions, the words associated with the familiar and unfamiliar group were grouped into eight and nine clusters, respectively. Under tactile conditions, the words were grouped into seven and 12 clusters, respectively. Similarly, the words were grouped into 14 and 16 clusters, respectively, under visual–tactile conditions. Each cluster was represented by the word with the highest suitable score within the cluster. However, since we prepared 10 test samples in this study, factor analysis calculates up to nine factors. Therefore, by repeating the cluster analysis, we reduced the number of conditions from more than nine to a maximum of nine.

Factor analysis

Factor analysis was performed using the words obtained from the cluster analysis. Prior to performing the factor analysis, the participants evaluated the representative words on a seven-point Likert scale, from –3 (do not feel at all) to 3 (strongly feel). The samples were presented in a random order to each subject. Figure 8 shows the evaluation results using the Likert method. A t-test (p = 0.05) was performed on the scores of the genuine and artificial leather samples with regard to each representative word; this test indicated that only a few words were observed to describe the qualities inherent to genuine or artificial leather. In other words, the samples displayed varying properties regardless of the type. Under visual conditions, two words associated with the familiar and unfamiliar groups were confirmed. In contrast, six words under tactile conditions corresponded. The results under visual conditions suggested a more significant effect on the impressions of leather as compared to that under tactile conditions.

Figure 8.
Results of the Likert method: (a) visual conditions; (b) tactile conditions; (c) visual–tactile conditions.

Finally, we performed a factor analysis (JMP 13, SAS Institute, Cary, NC, USA) using the representative words. Figure 9 summarizes the results of the factor analysis and the names of each factor. In some results, we deleted words under the following conditions: (a) when the word did not contribute to any factor or (b) when double loading was observed (a phenomenon where there are large loading values across multiple factors). Although experimental conditions were varied, we defined the same names for several factors because words with similar meanings were represented. Table 2 summarizes the descriptions concerning the factor name determination. The impression factors extracted under visual and visual–tactile conditions were similar. Therefore, it is suggested that the visual perception is more significant than its tactile perception.

Figure 9.
Results of factor analysis and names of the factors: (a) visual conditions; (b) tactile conditions; (c) visual–tactile conditions.

Table 2.
Summary of the descriptions of factor name determination

Condition Factor name Description of naming

Visual Surface shape Consisting of the words that describe the leather surface conditions

Impression of stateliness Consisting of the “profound,” which implies sensory depth and heaviness, and impressive appearance of surface ("texture")

Jet black Consisting of the words that describe the surface color and gloss of leather

Sturdiness Consisting of the words that describe the impression of toughness

Tactile Moisture Consisting of the words that describe “moist” and “soft”

Fineness Consisting of the words indicating both friction and smoothness

Surface shape Same reason as Visual conditions (familiar group)

Fineness Same reason as Tactile conditions (familiar group)

Visual–tactile Profound Consisting of the words indicating the surface sturdiness and impressive appearance of leather

Surface shape Same reason as Visual conditions (familiar group)

Jet black Same reason as Visual conditions (Unfamiliar group)

Dynamic gloss Consisting of the words relate to the spread of gloss when the participants touch the sample

Table 3 summarizes the changes in the number of words, clusters, and factors obtained from the subjective experiments discussed in this section.

Table 3.
Summary of the variation in the number of words, clusters, and factors under visual, tactile, and visual–tactile conditions

Experiment/Analysis Visual conditions
Tactile conditions
Visual–tactile conditions

Familiar group Unfamiliar group Familiar group Unfamiliar group Familiar group
Unfamiliar

group

Word extraction experiment 92 words 94 words 117 words

Suitable words selection experiment 24 words 28 words 13 words 31 words 28 words 29 words

Cluster analysis 8 clusters 9 clusters 7 clusters 9 clusters 9 clusters 9 clusters

Factor analysis 2 factors 2 factors 2 factors 2 factors 2 factors 2 factors

“Authenticity”

The leather “authenticity” was evaluated using Scheffe's paired comparison method. Under each condition, the same participants as in the Impression factors section participated. We chose pairs of test samples and presented them in a random order to the participants who evaluated them using seven levels, which ranged from “did not feel authentic at all (–3)” to “very strongly felt authentic (3).” The above-mentioned comparison was conducted for all possible combinations of samples (₁₀C₂ pairs). The results of the paired comparison method were scaled by a correspondence analysis that plots the data onto multiple dimensions. As a result, we obtained six (7 ranks − 1)-dimensional results. Table 4 presents the contribution rate of each dimension by the analysis. The contribution rate of each dimension is calculated by the following equation
$contribution rate = λ_{k} / \sum_{k^{'} = 1}^{K} λ_{k^{'}}$
(1)

Table 4.
Results of correspondence analysis (rounded to two decimal places)

Conditions Group D = dimensional results (contribution rate)

D1 D2 D3 D4 D5 D6

Visual Familiar 0.69 0.14 0.08 0.06 0.02 0.01

Unfamiliar 0.55 0.26 0.14 0.04 0.01 0.00

Tactile Familiar 0.54 0.31 0.09 0.04 0.01 0.00

Unfamiliar 0.51 0.21 0.16 0.07 0.03 0.03

Visual–tactile Familiar 0.77 0.13 0.09 0.01 0.00 0.00

Unfamiliar 0.73 0.15 0.05 0.05 0.02 0.00

Here, λ, k, and K are the eigenvalue, the dimension number, and the maximum dimension number, respectively. From Table 4, we confirmed that dimension 1's contribution rates under visual conditions (associated with the familiar group) and under visual–tactile conditions (for both groups) were particularly dominant. In other words, under these conditions, the results of the analysis can be almost entirely explained by dimension 1. Furthermore, as shown in Figure 10, we found that the seven-evaluation rank order plot (blue) is associated with dimension 1. From the above-mentioned results, we determined that dimension 1's score under visual conditions (in the familiar group) and under visual–tactile conditions (both groups) was the “authenticity” score. In contrast, the other conditions did not produce an “authenticity” score that is 1D. We assumed that this was due to the large evaluation variability within the participant group. In other words, under these conditions, there was no consistent perception of “authenticity” between the participants. Under tactile conditions where the participants relied on touch in an unrealistic setting, the evaluation variability between participants increased in both groups. For unfamiliar groups, scaling was possible under visual–tactile conditions. We believe that the variation in evaluation within the group was reduced due to the cross-modal phenomenon; in other words, the perception was informed by the interaction between different senses.

Figure 10.
Examples of results of the correspondence analysis. (Color online only.)

Figure 11 shows the “authenticity” score under each condition. In the graph, the vertical axis is the subjective score of “authenticity,” and the horizontal axis is the sample number. It was noted that sample no. 4 has a low “authenticity” score despite being genuine leather. This was considered related to the fact that some of the properties of sample no.4 differed from those of the other genuine leather samples in Figure 5 (for example, kurtosis value, irregularity component, and SMD).

Figure 11.
Scaled results of “authenticity.” We can confirm that the sample pairs connected with blue frames differ significantly (p < 0.05), while we cannot confirm that the samples inside the blue frame differ significantly. (Color online only.)

Quantification modeling

Based on the results of the fourth section, we attempted to quantify “authenticity” using the results obtained by the familiar group under visual conditions, as well as both groups under visual–tactile conditions. As shown in the hypothesis in Figure 1, we theorized that representative impressions could be estimated through the optimal integration of a certain number of the 59 leather properties listed in Table 1. Moreover, “authenticity” might be quantified by integrating these impressions. Accordingly, we proceeded to derive the estimated equations using multiple regression analyses. For the equations, we selected explanatory variables by minimizing the Akaike information criteria corrected (AICc) value.⁴⁴ A small AICc value indicates a superior estimation model in terms of robustness and data correlation. In this study, considering that the objective variable was the subjective score, we speculated that the estimation model derived from explanatory variables with small AICc values was a particularly relevant model for human perception of leather. However, we inferred that there is a high correlation between visual properties, despite differing measurement geometries, and therefore the subjective impressions may also be estimated using properties different from properties selected during modeling (for example, the 20°/45° L* and 30°/45° L* values). Considering the above, we discuss whether the selected visual properties are important factors in the formation of impressions. Here, the property values were converted to Z-scores and treated as variables.

Properties and impression factors

Figure 12 shows the high positive correlations between the subjective scores of the impression factors (horizontal axis) and the estimated scores of the properties (vertical axis) under visual conditions and visual–tactile conditions.

Figure 12.
Plots of the estimated scores versus the subjective scores: (a) visual conditions; (b) visual–tactile conditions.

Under visual conditions, the surface shape is described by the following equation
$surface shape = 0.76 \cdot x_{11} - 0.31 \cdot x_{12}$
(2)

Here, x₁₁ and x₁₂ represent the 60°/0° large grain component and the 45°/45° L* value, respectively. An R-squared value of 0.91 is obtained (p < 0.001). In addition, we found that the 60°/0° large grain component and the 45°/45° L* value highly correlated with the 45°/0° large grain component (R² = 0.91) and the 15°/0° irregularity component (R² = 0.67), respectively. These results show that the surface shape is mainly perceived as large grains in the region of the shade angle (i.e., the angle in the direction of diffuse reflection). Moreover, the large grains of the leather surface could not be observed when the specular L* value and the low-frequency value in the region of the highlight angle (i.e., the angle near the direction of specular reflection), which is related to L, were high. We therefore believe that the L value negatively affects the impression factor of the surface shape.

The impression of stateliness is described by the following equation
$impression of stateliness = 0.71 \cdot x_{21} + 0.24 \cdot x_{22} - 0.44 \cdot x_{23}$
(3)

Here, x₂₁, x₂₂, and x₂₃ are the 60°/0° irregularity component, the 15°/0° irregularity component, and the –15°/45° kurtosis value, respectively. An R-squared value of 0.96 is obtained (p < 0.001). In addition, we found that the 15°/0° irregularity component and the −15°/45° kurtosis value highly correlated with the 25°/0° irregularity component (R² = 0.77) and the 0°/45° kurtosis value (R² = 0.84), respectively. We did not find properties having high correlation with the 60°/0° irregularity component, indicating that surface irregularity at the shade angle is particularly important. These results show that the impression of stateliness is informed by the surface irregularity under multiple-observation angles. Also, the large kurtosis value at the shade angle indicates that the surface L* value is uniform, that is, no irregularity is detectable on the surface. The kurtosis value therefore negatively affects the impression.

Under visual–tactile conditions, the profound factor is described by the following equation
$profound = 0.74 \cdot x_{31} + 0.25 \cdot x_{32} - 1.56 \cdot x_{33}$
(4)

Here, x₃₁, x₃₂, and x₃₃ represent the return direction MMD, the 15°/0° grain size, and the maximum static friction coefficient, respectively. An R-squared value of 0.98 is obtained (p < 0.001). Based on these results, the profound factor is positively influenced by an irregularity in the dynamic friction coefficient and the grain size, and negatively influenced by the maximum static friction coefficient.

The surface shape is described by the following equation
$surface shape = 0.92 \cdot x_{41}$
(5)

Here, x₄₁ is the 15°/0° grain size. An R-squared value of 0.85 is obtained (p < 0.001). In addition, the 15°/0° grain size highly correlated with the 60°/0° grain size (R² = 0.81). Based on these results, the surface shape is positively influenced by the average grain size of the surface under the entire angle range. The impression may be estimated by the visual property alone.

The jet-black factor is described by the following equation
$jet black = - 0.43 \cdot y_{11} - 0.64 \cdot y_{12}$
(6)

Here, y₁₁ and y₁₂ represent the round-trip SMD and the 60°/0° large grain component, respectively. An R-squared value of 0.96 is obtained (p < 0.001). These results show that the jet-black factor is strongly perceived when no grains are visually and tactilely discernable on the surface.

Finally, the dynamic gloss is described by the following equation
$dynamic gloss = 1.40 \cdot y_{21} - 0.66 \cdot y_{22}$
(7)

Here, y₂₁ and y₂₂ represent the 15°/0° irregularity component and the 45°/45° L* value, respectively. An R-squared value of 0.88 is obtained (p < 0.01). Based on these results, the dynamic gloss is positively influenced by the surface irregularity and negatively influenced by the 45°/45° L* value. We believe that the participants visually perceived dynamic gloss by the change in gloss when touching the leather sample.

As described above, the representative leather impressions were quantified using visual and physical properties. Under visual–tactile conditions, half of the estimation equations reference only visual properties. In this study using leather, this result suggests that vision informs the human perception mechanism more significantly than tactile sensation. We considered that since the experiments were performed in a controlled environment with thin samples, the thermal properties and elasticity did not contribute significantly to the representative impressions.

Impression factors and “authenticity”

Figure 13 shows the high positive correlation between the subjective scores for “authenticity” (horizontal axis) and the estimated impressions scores (vertical axis).

Figure 13.
Plots of the estimated scores versus the subjective scores for “authenticity.”

The “authenticity” (associated with the familiar group) under visual conditions is described by the following equation
$“ authenticity ” (visual) = 0.59 \cdot z_{11} + 0.56 \cdot z_{12}$
(8)

Here, z₁₁ and z₁₂ are the surface shape and the impression of stateliness, respectively. An R-squared value of 0.80 is obtained (p < 0.001).

The “authenticity” (associated with the familiar group) under the visual–tactile conditions is described by the following equation
$“ authenticity ” (visual - tactile) = 0.53 \cdot z_{21} - 0.64 \cdot z_{22}$
(9)

Here, z₂₁ and z₂₂ are the profound and the surface shape under visual–tactile conditions, respectively. An R-squared value of 0.76 is obtained (p < 0.01).

Finally, “authenticity” (associated with the unfamiliar group) under visual–tactile conditions is described by the following equation
$“ authenticity (visual - tactile) = 0.72 \cdot z_{31}$
(10)

Here, z₃₁ is the jet black under visual–tactile conditions. An R-squared value of 0.52 is obtained (p < 0.05).

The relationship among “authenticity,” the representative impressions, and the leather properties is illustrated in the subjective response hierarchy, shown in Figure 14. From the above, “authenticity” under each condition could be quantified in terms of the leather properties through each impression factor. We concluded that visual information plays a major role in the perception of leather “authenticity,” and perceptual variability between participants who are unfamiliar with leather can be reduced by allowing the participants to use more sensors (senses).

Figure 14.
Hierarchy of subjective responses. MMD: variation of average dynamic friction coefficient; SMD: average deviation of surface roughness.

Verification experiment

We derived perception models of the “authenticity” of leather. However, the models are based on only 10 test samples. Therefore, we conducted a verification experiment with an increased number of samples and participants to confirm the robustness of the models.

Figure 15 shows the enlarged images of the additional 12 leather samples. No. 11–16 are images of artificial leather samples and no. 17–22 are images of genuine leather samples. The samples were black and had dimensions of 210 mm × 300 mm. These RGB images were captured under D65 diffuse illumination.

Figure 15.
Enlarged images of the additional samples.

We performed measurement experiments and subjective experiments using these additional samples. The experimental conditions (environment and measurement systems) were the same as in the previous experiments. In the subjective experiment, the familiar group consisted of 15 participants, including an additional five people (four men and one woman), 20–50 years of age. The unfamiliar group consisted of 15 participants, the same as the previous subjective experiments.

Figure 16 shows the correlation between the subjective scores for “authenticity” (horizontal axis) and the estimated scores for “authenticity” (vertical axis) of the additional samples. The black circles represent the 10 original test samples and the colored polygons represent the additional samples, whereas the black dotted lines indicate the 95% prediction interval of the models estimated from the test samples. From this figure, under the visual and visual–tactile conditions of the familiar group, all estimates for the additional samples were within the 95% prediction interval. Therefore, we could confirm the robustness of the proposed models, whereas under visual–tactile conditions of the unfamiliar group, only sample no. 20 fell outside the predicted interval. It is was suggested that there are other properties that affect perception besides the properties we studied. Although the participants in the unfamiliar group used multi-sensors to perceive “authenticity,” the evaluation variability was still large compared to the familiar group. However, since the other 11 samples were within the prediction interval, we determined that the model of the unfamiliar group demonstrated relative robustness as well.

Figure 16.
Results of the verification experiment. (Color online only.)

Limitations

Although we confirmed the robustness of the “authenticity” estimation model, the following limitations were noted. The participants were all men belonging to a single nationality. In this study, data size under controlled conditions was comparable to conventional studies (for example, Chen et al.,²² Okamoto et al.,²³,²⁵ Nagano et al.,²⁴ and Schiffman³⁰), and statistically correlated results were obtained even with the number of participants and samples considered in this study. However, gender and cultural differences in perception were fully assumed. Therefore, to derive a generalized model, it is necessary to conduct more practical verification by increasing the number of participants.

Furthermore, only black cowhide was used for genuine leather samples. However, surface properties may vary depending on the animal type. In addition, surface color information has an important effect on perception. In the future, we will conduct experiments using colored and other animal leather samples. Also, although this study focused on only tactile touch, the perception of grasping and holding may be related to “authenticity.”

Conclusions

In this study, to quantitatively evaluate shitsukan ambiguity perceived by people, we constructed hierarchical subjective responses, and derived the estimation of leather “authenticity” under visual, tactile, and visual–tactile conditions using the visual and physical properties.

The derived estimation models highly correlated with subjective scores (see Figures 13 and 14). However, since the properties related to perception differed depending on the attributes of the group, it is necessary to adjust the concerned properties in artificial leather production according to the target consumer (that is, consumers who already owned leather products or those who were purchasing leather products for the first time).

Although we were unable to quantify “authenticity” perceived by the unfamiliar group under visual and tactile conditions, it was feasible under visual–tactile conditions (see Table 3). This result indicated that the evaluation variability within the group was somehow reduced and suggested the influence of a cross-modal phenomenon (as opposed to a multi-modal phenomenon), which is a synergistic effect produced by multi-sensory perception, that is, when the participants evaluated leather through more of their senses.

The results of this study can be used to develop attractive artificial leather, which is expected to contribute to the preservation of animal rights while promoting the sale of “leather” products. The obtained results also elucidate the mechanism of perceiving leather.

(a)Method	Properties
Color property measurement(5 geometries)	- Lab* value- Skewness and kurtosis of L*
Surface characterization(4 geometries)	- Average grain size- Integrated value of spatial frequency band (irregularity, large grain, and small grain components) - Integrated value of the entire spatial frequency range
(b)Device	Properties
KES-FB4-A Surface Tester	Values obtained for the forward, return, and round trip1. MIU (average dynamic friction coefficient)2. MMD (variation of average dynamic friction coefficient) 3. SMD (average deviation of surface roughness) - Maximum static friction coefficient
KES-F7 Thermo Labo	- q_max value (peak heat flux)
KES-FB3-A Compression Tester	- LC (compression linearity) - WC (compression energy) - RC (compression recoverability)

Condition	Factor name	Description of naming
Visual	Surface shape	Consisting of the words that describe the leather surface conditions
Impression of stateliness	Consisting of the “profound,” which implies sensory depth and heaviness, and impressive appearance of surface ("texture")
Jet black	Consisting of the words that describe the surface color and gloss of leather
Sturdiness	Consisting of the words that describe the impression of toughness
Tactile	Moisture	Consisting of the words that describe “moist” and “soft”
Fineness	Consisting of the words indicating both friction and smoothness
Surface shape	Same reason as Visual conditions (familiar group)
Fineness	Same reason as Tactile conditions (familiar group)
Visual–tactile	Profound	Consisting of the words indicating the surface sturdiness and impressive appearance of leather
Surface shape	Same reason as Visual conditions (familiar group)
Jet black	Same reason as Visual conditions (Unfamiliar group)
Dynamic gloss	Consisting of the words relate to the spread of gloss when the participants touch the sample

Experiment/Analysis	Visual conditions	Tactile conditions	Visual–tactile conditions
Word extraction experiment	92 words	94 words	117 words
Suitable words selection experiment	24 words	28 words	13 words	31 words	28 words	29 words
Cluster analysis	8 clusters	9 clusters	7 clusters	9 clusters	9 clusters	9 clusters
Factor analysis	2 factors	2 factors	2 factors	2 factors	2 factors	2 factors

Conditions	Group	D = dimensional results (contribution rate)
Visual	Familiar	0.69	0.14	0.08	0.06	0.02	0.01
Unfamiliar	0.55	0.26	0.14	0.04	0.01	0.00
Tactile	Familiar	0.54	0.31	0.09	0.04	0.01	0.00
Unfamiliar	0.51	0.21	0.16	0.07	0.03	0.03
Visual–tactile	Familiar	0.77	0.13	0.09	0.01	0.00	0.00
Unfamiliar	0.73	0.15	0.05	0.05	0.02	0.00

Footnotes

Declaration of conflicting interests

The authors have no conflicts of interest to declare.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Shuhei Watanabe

References

Komatsu

Goda

Neural mechanisms of material perception: quest on shitsukan. Neuroscience 2018; 392: 329–347.

Spence

Shitsukan - the multisensory perception of quality. Multisens Res 2020; 33: 737–775.

Hirai

Horiuchi

Investigation of Japanese onomatopoeias as features for SHITSUKAN-based image retrieval. In: 2013 IEEE international symposium on multimedia, Anaheim, California, USA, 9–11 December 2013, pp.399–400. New York: IEEE.

Horiuchi

Challenges in shitsukan imaging. In: the 4^th Asia color association conference, Chiang Mai, Thailand, 5–8 December 2018, pp.30–32. Color society of Thailand.

Citrin

Stem

Jr Spangenberg

, et al. Consumer need for tactile input: an internet retailing challenge. J Bus Res 2003; 56: 915–922.

Brown McCabe

Nowlis

SM.

The effect of examining actual products or product descriptions on consumer preference. J Consum Psychol 2003; 13: 431–439.

Fleming

Nishida

Gegenfurtner

KR.

Perception of material properties (part I). Vision Res 2015; 109: 123–236.

Fleming

Nishida

Gegenfurtner

KR.

Perception of material properties (part II). Vision Res 2015; 115: 157–302.

Fujisaki

Tokita

Kariya

Perception of the material properties of wood based on vision, audition, and touch. Vision Res 2015; 109: 185–200.

10.

Matsuoka

Kanai

Tsuji

, et al. Predicting texture image of covering fabric for car seat by physical properties. J Text Eng 2008; 54: 63–74.

11.

Watanabe

One-shot multi-angle measurement device for evaluating the sparkle impression. J. Imaging Sci Technol 2019; 63: 060401-1–060401-8.

12.

Ferrero

Bayón

The measurement of sparkle. Metrologia 2015; 52: 317–323.

13.

Beghetto

Ugolini

Tortora

, et al . Green biopolymers for ecofriendly leather. J Mater Sci Eng A 2017; 7: 199–205.

14.

Plannthin

DK.

Animal ethics and welfare in the fashion and lifestyle industries. In: Muthu

Gardetti

(eds) Green fashion environmental footprints and eco-design of products and processes. Vol 2. Singapore: Springer, 2016, pp.49–122.

15.

Tani

Nagai

Koida

, et al. Experts and novices use the same factors-but differently-to evaluate pearl quality. PLoS ONE 2014; 9: e86400.

16.

Watanabe

Tominaga

, and Horiuchi

The difference in impression between genuine and artificial leather: quantifying the feeling of authenticity. J Percept Imaging 2020; 3: 1–11.

17.

Watanabe

Horiuchi

Modeling perceptions using common impressions: perceptual “authenticity”, “luxury”, and “quaintness” for leather. Text Res J. Epub ahead of print 23 June 2020. DOI: 10.1177/0040517520935534.

18.

Goodman

Montgomery

Bialek

, et al. Measurement of naturalness (MONAT). In: 12th IMEKO TC1-TC7 joint symposium on man, science & measurement 2008, Annecy, France, 3–5 September 2008. International Measurement Confederation: Budapest, Hungary, pp.31–36.

19.

Bialek

Forbes

Goodman

, et al. Model development to predict perceived degree of naturalness. In: 19th IMEKO world congress fundamental and applied metrology, Lisbon, Portugal, 6–11 September 2009. International Measurement Confederation: Budapest, Hungary, pp.2124–2129.

20.

Overvliet

Karana

Soto-Faraco

Perception of naturalness in textiles. Mater Des 2016; 90: 1192–1199.

21.

Hashim

IHM

Kumamoto

Takemura

, et al. Tactile evaluation feedback system for multi-layered structure inspired by human tactile perception mechanism. Sensors (Basel) 2017; 17: 2601.

22.

Chen

Barnes

Childs

THC

, et al. Materials’ tactile testing and characterisation for consumer products’ affective packaging design. Mater Des 2009; 30: 4299–4310.

23.

Okamoto

Nagano

Kidoma

, et al. Specification of individuality in causal relationships among texture-related attributes, emotions, and preferences. Int J Affective Eng 2016; 15: 11–19.

24.

Nagano

Okamoto

Yamada

Semantically layered structure of tactile textures. In: haptics: neuroscience, devices, modeling, and applications (EuroHaptics 2014) (eds Auvray M and Duriez Z), Versailles, France, 24–26 June 2014, vol 8618. Berlin: Springer.

25.

Okamoto

Kojima

Yamagishi

, et al. Layered-modeling of affective and sensory experiences using structural equation modeling: touch experiences of plastic surfaces as an example. IEEE Trans Affect Comput. Epub ahead of print 7 November 2018. DOI: 10.1109/TAFFC.2018.2879944.

26.

Kawabata

Niwa

Wang

Objective hand measurement of nonwoven fabrics part I: development of the equations. Text Res J 1994; 64: 597–610.

27.

Kawabata

Mori

Niwa

An experiment on human sensory measurement and its objective measurement case of the measurement of seam pucker level. Int J Cloth Sci Tech 1997; 9: 203–206.

28.

Kawabata

Niwa

Clothing engineering based on objective measurement technology. Int J Cloth Sci Tech 1998; 10: 263–272.

29.

Kawabata

Niwa

Yamashita

Recent developments in the evaluation technology of fiber and textiles: toward the engineered design of textile performance. J Appl Polym Sci 2002; 83: 687–702.

30.

Schiffman

HR.

Sensation and perception: an integrated approach. 5th ed. New York: John Wiley & Sons, 2001, pp.2–3.

31.

Hui

TKL

Sherratt

RS.

Towards disappearing user interfaces for ubiquitous computing: human enhancement from sixth sense to super senses. J Ambient Intell Humaniz Comput 2017; 8: 449–465.

32.

Hiramatsu

Goda

Komatsu

Transformation from image-based to perceptual representation of materials along the human ventral visual pathway. Neuroimage 2011; 57: 482–494.

33.

Goda

Tachibana

Okazawa

, et al. Representation of the material properties of objects in the visual cortex of nonhuman primates. J Neurosci 2014; 34: 2660–2673.

34.

Bergmann Tiest

WM.

Tactual perception of material properties. Vision Res 2010; 50: 2775–2782.

35.

Manita

Suzuki

Homma

, et al. A top-down cortical circuit for accurate sensory perception. Neuron 2015; 86: 1304–1316.

36.

Biocca

Kim

Choi

Visual touch in virtual environments: An exploratory study of presence, multimodal interfaces, and cross-modal sensory illusions. Presence-Teleop Virt 2001; 10: 247–265.

37.

ISO 11664-4/CIE S 014-4:1976. Colorimetry – part 4: CIE 1976 L*a*b* Colour spaces.

38.

Dooley

Shaw

Noise perception in electrophotography. J Appl Photogr Engng 1979; 5: 190–196.

39.

Bajaj

Bains

Evaluation of tactile comfort of mulberry silk waste and viscose blended knitted fabrics by using Kawabata Evaluation System (KES) for apparel use. Curr J Appl Sci Technol 2019; 38: 1–15.

40.

Capdevila

Carrera-Gallissà

Escusa

, et al. Canonical analysis of the Kawabata and sliding fabric friction measurement methods. J Text I 2020; 111: 890–896.

41.

Yang

Wang

Innovative artificial leather with high thermal conductivity as a new leather product. Text Res J 2017; 87: 2487–2504.

42.

Roh

Kim

SH.

Classification of synthetic polyurethane leather by mechanical properties according to consumers’ preference for fashion items. Fiber Polym 2013; 14: 1731–1738.

43.

Okamoto

Nagano

Yamada

Psychophysical dimensions of tactile perception of textures. IEEE Trans Haptics 2013; 6: 81–93.

44.

Akaike

Information theory and an extension of the maximum likelihood principle. In: Petrov BN and Csáki F (eds) 2nd international symposium on information theory, Tsahkadsor, Armenia, 2–8 September 1971, pp.267–281. Budapest: Akadémiai Kiadó.

Experiment/Analysis	Visual conditions		Tactile conditions		Visual–tactile conditions
Experiment/Analysis	Familiar group	Unfamiliar group	Familiar group	Unfamiliar group	Familiar group	Unfamiliar group
Word extraction experiment	92 words		94 words		117 words
Suitable words selection experiment	24 words	28 words	13 words	31 words	28 words	29 words
Cluster analysis	8 clusters	9 clusters	7 clusters	9 clusters	9 clusters	9 clusters
Factor analysis	2 factors	2 factors	2 factors	2 factors	2 factors	2 factors