Methods of 3D data applications to inform design decisions for physical comfort

2.1 Definitions of comfort –comfort in the literature

A debate lies in the body of scientific literature, regarding a commonly accepted definition of comfort. Slater [5] defines comfort as a pleasant state of physiological, psychological and physical harmony between a human being and the environment. Richards [6] states comfort is the state of mind that involves a sense of subjective well-being in reaction to an environment or a situation. Vink [7] focuses on the subjective view of comfort by stating that in the context of air flights “Comfort is a subjective experience. For Passenger 1 on a long distance flight, back discomfort is of great importance. Passenger 2 wants a reduction in noise and Passenger 3 needs more space.” According to Vink [7], comfort is defined as (1) a construct subjectively defined, (2) it is a reaction to the environment and (3) it is affected by factors of various natures linking to the participant (physical, psychological and physiological). This paper focuses primarily upon the physical properties of comfort i.e., the effect of the physical interaction between a human and a product over limited interaction time (the observed interaction lasts for 3–5 minutes).

2.2 Physical loading, an influencing factor of comfort

Among the many comfort studies which link discomfort to physical loading, Kee et al. [8] investigated the relationships between the external load on human body (subjective measures of discomfort) and objective measures related to the assessment of postural stresses based on literature survey. Objective measures included posture holding time, Maximum Holding Time (MHT), torque at joints, Lifting Index (LI) and compressive force (CF) at L5/S1. The major relationships identified in this literature survey were the following: 1) postural discomfort linearly increased with increasing holding time, and holding force, 2) whole body discomfort was inversely linearly proportional to the MHT, 3) body-part discomfort was related to objective measures such as torque at the relevant joint and 4) discomfort was strongly linearly related to LIs and CFs, and Thus, discomfort might be used as a measure for quantifying postural stresses.

In a similar context, Zenk et al. [9] conducted an objective assessment approach which evaluates the concept of “optimal load distribution”, based on the identification of a close relationship between the pressure on the seat and the discomfort felt by the person sitting. The results indicate that in the seat position with the pressure distribution corresponding to the most comfortable posture the pressure in the intervertebral disc is lowest.

Hence, if the physical load forced on the user is kept to low levels and the exposure of users to the load lasts for a considerable amount of time then their response can relate to objective measures. There is a strong connection between discomfort and physical dimensions, such as postural stress.

2.3 Past research on collection of 2D and 3D human data with focus on ear data and improving design

Anthropometry is considered the very ergonomic core of any attempt to resolve the dilemma of fitting the task to the human [10]. In regards to external ear products such as Bluetooth headsets and headphones, good fit is a crucial physical factor of comfort to ensure the success of these products. Designers require anthropometric data to identify human factors and inform design decisions with respect to external ear devices. Current approaches are restricted tithe presentation of anthropometric data only. The collection of ear data includes the use of various measurement instruments. Jung and Jung [2] provided anthropometric dimensions of ears of Korean subjects using digital callipers. Other methods suggest the use of simple geometric calculations to acquire dimensions from a 2D photograph by setting reference points before taking the photographs. In regards to data collection of other body parts, such as head and legs data, other, relatively non-invasive, 3D imaging techniques are applied. These include various forms of stereophotogrammetry [11], topography techniques [12] and surface scanning technologies [13]. In their paper, Stavrakos & Ahmed-Kristensen [14] compared three methods of collection of ear data (with a digital calliper, photogrammetrtic method and 3D scanning method), in terms of cost, time, accuracy and comfort, they concluded that the 3D scanning method is the most costly yet the most accurateof all.

In other disciplines such as seating design, design engineers have attempted to design desks and chairs based on anthropometric data [1, 3]. Parcells et al. [3] studied the mismatch between furniture and students’ dimensions by measuring anthropometric characteristics of American children aged 11–13 years and the dimensions of their classrooms’ desks and chairs, reporting that only 18.9% of students could find an appropriate match [15]. Other studies provide detailed anthropometric data and some of them also offer recommendations for design [16]. In their study, Stavrakos & Ahmed-Kristensen et al. [17] presented a validated methodological framework of how to generate archetypes from an anthropometric dataset of ears measured manually with a Vernier caliper, in order to build test panels for the execution of more reliable comfort studies.

The majority of these studies focused on the acquirement of anthropometric data, and this was limited to physical match. The anthropometric dimensions presented in these papers were not further analyzed from a mathematical point of view to acquire a more advanced knowledge of the human geometry. Furthermore, no studies were found that proposed methodologies to predict human factors with the attempt to improve physical comfort factors in product design. In particular in the ear industry, there appeared to be a gap in the definition of a validated methodological framework to link ear anthropometry to design.

2.4 Research questions and hypothesis

This section presents the main research question (RQ) and hypothesis of the research presented. The main research question was:

How can 3D human data support the design of products – in this case, external, ear products for short term human – product interactions? The set of hypothesis linked to the research question are presented here:

(H1) By meaningful manipulations of 3D human data, it is possible to benchmark human dimensions against product dimensions with the aim of supporting product design, and

In addition to the hypothesis presented, in this paper, the archetype methodology by Stavrakos and Ahmed-Kristensen et al. [17] was reproduced using a 3Dimensional dataset. Hence, the archetype methodology will only be mentioned briefly in the Methods section.

3.1 Early product understanding

This paper investigates short-term interactions(2-3 minutes) between humans and external – ear worn products, such as Bluetooth headsets, headphones, etc. Initially, product categories are created based on wearing style and fit. Five categories of external-ear worn products were defined (see Fig. 2) for this study. At this stage, there was also the possibility of defining critical product components that touch critical areas of the human body. The interfaces would require a different treatment from a data collection point of view, such as the ear gel component in the “in the ear headsets” product category, which touches upon the in-the-ear area. However, this paper focuses on is the external ear area and paper does not focus on productcomponents as the main anthropometric area. The researchers of this paper aimed to eliminate the possibility of having to treat in-the-ear data which would require different scanning techniques. The products mentioned in this paper are objects seen as one object with stabile physical properties.

In the next step, critical areas were defined on the human surface. Critical areas were defined as the areas where a product is in physical contact to the human body.

Once these areas were defined on the human head and ears, critical human dimensions were defined to describe as accurately as possible the critical area from a human-product interaction point of view, rather than a geometrical point of view, e.g. the ear to ear – over the head top dimension. The associations between critical product areas and the determination of critical human dimensions, for product families 1 and 3 can be seen in Fig. 3.

Once the associations between critical areas of human-product interaction and human dimensions were completed, the authors issued a table listing all critical human dimensions to be collected in order to achieve understanding of the product and the human body. In total, 25 linear and non-linear human dimensions were identified in respect to the 5 product categories of ear-worn products.

3.2 3-Dimensional data collection

Once the critical human dimensions were determined, head and ear data were collected. The various steps of this process (see Fig. 1) comprise of (1) a selection of the population, (2) the definition of a representative user sample, and (3) the scanning of the population. Table 1 also shows the demographics of the data collection of this study and the measurements performed on the 3D head scans to collect all critical human dimensions.

To ensure that the user sample was representative, a population with homogenous, ethnographic properties should be selected. The population of a country, which in this instance was Denmark, was selected for the execution of this study. To ensure that the user sample chosen represents Denmark’s population -which is approximately 5,500,000 persons-, the statistical sample size was calculated using the following mathematical equation: Necessary Sample Size = ( Z - score ) 2 * StdDev * ( 1 - StdDev ) / ( margin of error ) 2 In the case of Denmark, the sample size required is 196 people for a confidence level of 95%, confidence interval = 7 and population size = 5,500.00. This means that if approximately 200 persons are randomly selected out of the 5,500,000, these can statistically sufficiently represent the population at hand from an anthropometric point of view. Hence, a randomized sample of 200 Danish people (100 men, 100 women) was chosen with ages ranging from 22 to 67 years to match the requirements of the calculated sample size for the scanning of the persons.

The steps of the scanning process are briefly depicted in Fig. 4. The 196 persons were scanned using a system consisting of one projector and two cameras which were both calibrated. A structured design was employed where patterns were projected onto the object’s surface, pixel values were assigned by calculating the exact distance between points on the object’s surface and the focal plane of the camera. In this system all visible points were calculated simultaneously in a single capture in a few seconds. Once the scanning took place the two images produced by the two cameras were merged with the help of 3D modelling software. The outcome of this was a single 3D model.

The 196 3D Models were later processed using a 3D imaging software (G.O.M. Inspect 2.0). The purpose of this iteration was to refine critical areas which were hard to scan on site, such as areas behind the ears and human surface covered with hair. The G.O.M. imaging software allowed these areas to be restored by making use of a predictive algorithm that generated the missing areas by using the data of neighbouring human surface.

3.3 Applications of 3-dimensional data

The various steps of this process are shown in Fig. 1. Once the scanning of the participants and the refinement of the scans were complete, a dataset of 196 human head scans were available. Three main manipulations were identified and performed in this paper. They are the Basic and In-Depth benchmarking between Products and Human Factors, the retrieval of critical dimensions through Qualitative Methods of Data Analysis (e.g. Visualizations of 3D-Data), and the Creation of Focus Groups by utilizing the archetype methodology described in depth in Stavrakos & Ahmed-Kristensen [17] and summarized for this paper in Section 3.6.

3.4 Basic and in-depth benchmarking between products and human factors

Regarding the basic benchmarking, the main step of this process (see Fig. 1) concerns the selection of reference points and generation of a matrix of critical, linear human dimensions. In the basic benchmarking, the matrix of the associations between critical product areas and human dimensions was used (See Fig. 3). As stated earlier, 25 linear and non-linear human dimensions were generated from the matrix. The 25 dimensions were identified on all 196 head scans by generating reference points for the linear dimensions (starting point – ending point) and by drawing the respective curves to reproduce the non-linear ones. After all reference points were applied, the dimensions between the starting and ending point for each one of the linear dimensions were calculated, as well as the non-linear ones (see Fig. 5). At this stage, 4,900 measurements were retrieved manually from the imaging software. All activities were performed in G.O.M. Inspect. The measurements were inserted into a matrix and several statistical values were calculated, such as, minimum (MIN), maximum (MAX), median (MN) and average (AVE) values. Furthermore, tables of percentiles and frequencydiagrams were calculated for all dimensions. These are presented in the results section.

The in-depth benchmarking (see Fig. 1) comprises of (1) a mathematical modelling of products, (2) selection of critical angles and through-the-mesh dimensions and (3) application of geometrical elements on the 196 scans. Considerations were made regarding the generation of additional human dimensions which could be retrieved by working through the mesh of points of the 3D scans, compared to the 25 dimensions which were identified on the surface of the human heads. For this purpose, 5 additional, through-the-mesh dimensions were defined. These can be seen in Fig. 6. The first two dimensions, Back Canal Top and Back Canal Bottom connect the top and bottom point of the ear canal to the middle point behind the ear. These dimensions are useful for the design of ear-product categories 4 and 5 (see Fig. 2), which covers ear products worn behind the ear and products that includeear hooks.

Calculating through-the-mesh dimensions, offered the possibility of generating three crucial dimensions for the design of headphones and products worn around the head. Figure 6 presents head height, head breadth top and head breadth bottom, which can inform the design of product categories 1 and 2, i.e., headphones over and around the ears, respectively. Two more dimensions were also added to support product categories 1 and 2, i.e., the design of headphones. These were angles φ1 and φ2, as shown in Fig. 7. The angle φ1 is calculated as the angle formed (see Fig. 7) between segment (BC) and the height of the trapeze (ABCD), which, in this case, is segment L3. Segments (AB) and (DC), which are contained in the trapeze, are the human dimensions Head Breadth Top and Head Breadth Bottom, respectively, which were identified earlier (see Fig. 6):

( tan φ ) - 1 = ( ( AB ) - ( DC ) ) / L 3

For the calculation of φ2, two planes were formed. Plane A was formed by the two ear connection points top and bottom, and point F, as shown in Fig. 7. Plane B was formed by the same ear connection points and the middle point along back circumference, i.e., the behind the ear middle point, as shownin Fig. 6.

The angle φ2 was calculated in G.O.M. Inspect is the angle between planes A and B for both left and right ear. Both angles were calculated for all 196 head scans. In total, 9 more dimensions were added to the list of critical human dimensions (Back Ear Canal Top and Bottom for both right and left ear, Head height, Head Breadth Top, Head Breadth Bottom and angles φ1 and φ2), i.e., 1,764 additional measurements were executed.

3.5 Qualitative methods of data analysis – visualisations of data

The visualization of data (see Fig. 1) comprises of (1) the selection of reference points and curved lines, (2) overlay of geometry on the 196 scans, (3) and the calculation of product dimensions on the final image of the superimposed curves. An additional way to manipulate 3D data is to inform designers of the shape of product components and to generate human dimensions based on a qualitative understanding of human geometry. In this paper, additional human dimensions were generated through visualization of 3D human data. Figure 7 shows a Bluetooth headset worn behind the ear. Initially two interfaces were identified, the one resting over the ear (green colour) and the other (red colour) resting behind the ear. An advanced knowledge of the area behind the ear is needed to inform the dimensions of the back product component.

Additionally, a joint movement, in the position where point A lies in Fig. 7, was identified with 3 degrees of freedom (dF = 3). The mechanical model of the product components indicates that the back component rotates around point A, which lies on the frontal part of the ear area.

Visualisation of the back ear area was executed with respect to point A. The back component of the Bluetooth device was benchmarked against the back ear circumference which was calculated for all 196 headscans as one of the 25 ear dimensions identified during the Basic Benchmarking phase. The 196 curves were overlayed together in the same chart, maintaining a steady, reference point for all curves, which was the Ear Connection Point B, as shownin Fig. 8.

Point A, linking to the respective point on the product, was identified earlier (see Fig. 7), as well as points C (middle point, behind the ear) and D (lower back ear) were also determined on all 196 head scans. Segments (AB) and (BC) were calculated for all head scans. The geometry of the 196 curves is also available to contribute in the shaping of the back component part (see 4.1.2 in the Findingssection).

3.6 The issuing of focus groups with the help of 3D data analysis methods

Another meaningful way to apply the 3-dimensional dataset is to generate archetype persons out of the 196 participants, that is, define representative persons of the head and ear geometry with the aim of creating a representative user panel to use in the various comfort studies performed during the early and later conceptual phase of the product. The methodology of generating archetypes is described thoroughly in [17]. This section will briefly describe the archetype methodology.

A series of critical ear dimensions were assigned in the issuing of the five product families, as seen in Figs. 2 and 3. Cluster analysis was performed on the 196 persons in all 5 product categories, selecting for each one of the five categories, the critical ear and head dimensions associated with each product category. In the example of product category 3 (see Fig. 3), the three critical ear dimensions were the Concha X, Concha Y and Inner Circumference. The data was clustered using the Ward’s minimum variance method. The 196 participants were clustered based on the three selected ear dimensions. Hence, the data of ears was clustered in 9 meaningful groups. Figure 9 shows an example of how 2 out of the 9 clusters were formed.

As shown in Fig. 9, each cluster contained roughly 15–25 participants (in Fig. 9, cluster 1 contained 24 persons). In the next step frequency diagrams were made for each ear dimension in each group as shown in Fig. 10 for Concha X, Group 1.

This resulted in 3 frequency diagrams for each of the three ear dimensions within each group. Popular intervals were chosen to include at least 60% of the participants for each group. In the case of Concha X in Fig. 10 these intervals were interval 6 and 7. The archetype person selected based on the individual who belonged in all three popular intervals for all the three dimensions. The archetypes were then used by combining their scans and dimensions and also identify the person for the focus group to collect qualitative feedback.

4.1 Applications of 3-dimensional data

4.2 Discussion and implications

This section presents a discussion regarding the phases of the methodological framework and other aspects regarding the applications of 3-dimensional data which were presented earlier during the Findings section.