Research>Digital Human for Human Centered Design

Well Fitting Eyeglass Frames for Japanese Faces

>> Japanese

In a collaboration study with the Charmant Group/Horikawa Inc., the Digital Human Research Center analyzed the variations of the 3-D face shapes of 56 Japanese adult males in order to improve the morphological fit of eyeglass frames. As a result, it was found that the existing method of size classification for eyeglass frames did not conform to the actual individual facial differences. Horikawa Inc. classified faces into four types and designed new eyeglass frames accordingly. In addition to a 3-D morphological fit, this new frame offers a level of comfort which has never been experienced before. The new fit feels as if it is wrapping around the back of one's ears, not like the tight feeling a person gets on the sides of the face with existing frames. This frame has been sold by Charmant Group/Charmant Inc. as the "Co-Co-Chi" brand and has been targeted for seniors since July, 2001.

Distribution map of Japanese faces
Research Background

Traditional eyeglass frames were designed using face dimensions. Size classifications were also based on face dimensions. In addition, final custom fitting of eyeglass frames onto individual faces was done by retailers. For these reasons, while these eyeglasses were easily adjustable, most of them had a tendency to require re-adjustment at some point. Moreover, many complaints about discomfort around the ears and nose were received. Horikawa Inc. decided to develop a new frame that is more comfortable than traditional frames, and fits properly from the beginning with little adjusting needed.

Individual Differences on Three-Dimensional Face Forms

It was unknown what features were related to significant individual differences between 3-D face shapes. Therefore, it was decided to analyze 3-D facial forms. We measured 56 volunteer Japanese adult males.


In designing eyeglass frames the shapes behind the ears need to be considered. There were no devices available to measure such shapes, therefore plaster casts of the facial forms (from 3cm above the glabella to under the nose) of 56 volunteers were made. The positions of 29 anatomical landmarks were marked using water based pen before making the plaster casts, and, by doing this, plaster casts were made with markings transferred from the landmarks.

A Digital Model of a Face

For each subject, a homologous model of a face, composed of 211 data points based on the locations of the landmarks, was created.

Calculation of a Distribution Map

Using this model, the distance between the 2 shapes are defined. The distance matrix is made by calculating the distances for all of the 56 subjects' combinations. By analyzing the distance matrix using multi-dimensional scaling, variable axes that show individual differences and the points of each individual on the axis can be calculated. Here, 4 variable axes accounted for approximately 90% of the variations in facial forms.

The 4 axes are not correlated to each other. Data is extracted so that the first axis has the most information and the second has the next greatest amount of information. A distribution map of 3-D facial shapes was drawn using the first and second axes. The third and fourth axes represent the individual differences that are not directly related to the design of eyeglass frames. The first 2 axes explain approximately 80% of the total variations in individual differences. Also, the points of each axis show a normal distribution.

In order to simplify the interpretation of the first and second axes, the virtual shapes that are located at the edge of the distribution are illustrated for each axis on the distribution map. The first axis is the size factor for a face and is mainly related to facial depth. The second axis is the breadth factor of a face. The second axis is also related to the depth of the set of the eyes. These variable axes cannot be seen in analysis based on traditional measurements.

Distribution map of subjects based on the first 2 scales including average and average +-3S.D. forms
Facial Type Classification and Eyeglasses for Each Type

Horikawa, Inc. aimed at covering as wide of a range as possible with the fewest number of frame types. The company believed that four types would be able to cover most individual differences. Based on the distribution map, the 56 subjects were divided into 4 groups.

Distribution map of faces and their grouping
Distribution map of faces and their grouping

Next, the mean shape of each group was calculated using the technology developed by the Digital Human Research Center. In the order of smaller interpupillary distance to greater distance, it goes: type A (small face), type B (narrow face), type C (big face), and type D (wide face). Horikawa, Inc. suggested making a frame that offers a level of comfort which has never been experienced before. The company designed four types of eyeglass frames, one for each group, based on the mean shape of each group. The new fit feels as if it is wrapping around the back of one's ears, not like the tight feeling a person gets on the sides of the face with existing frames.

A new eyeglass frame and a traditional eyeglass frame
A new eyeglass frame (left) and a traditional eyeglass frame (right)
Evaluation of the Fit

An evaluation experiment was conducted at the Digital Human Research Center for the fit of the eyeglass frame designs based on the new concept. The evaluation was to specifically find if the newly proposed fit, "not tight, but non-slipping," was realized and accepted by consumers. The evaluation method consisted of sensory evaluation and physical measurements - the tightening force of the frame and the amount of slippage when the subjects shook their heads.

38 Japanese male volunteers participated in this evaluation experiment. A measurement-based formula was sought that enables estimation of a location on a distribution map. The subject can then be typed based on the resultant location on the distribution map. From the estimated location, we determined the most compatible frame among the four frames and named it frame 1. The frame that is considered to have the second highest compatibility became frame 2. The frame considered to have the least compatibility was frame 4. Each participant [types A to D] has a different facial shape, therefore, the frame which is considered to be frame 1 varies among participants. In addition to the 4 new frames, we also provided a traditional frame of similar appearance and weight. The participants, therefore, tried a total of 5 frames.

press and slip

The diagram shows the result of the physical measurements of type B (narrow) faces. The horizontal axis represents the tightness of the frame and the vertical axis represents the amount of slippage when the head was shaken. The data located at the lower right shows a frame that does not slip since the tightening force is high. Although it is not shown on this chart, traditional frames belong to this. The data located upper left shows less tightening force and high frame slippage. The frame 1, shown as , did not slip although the tightening force is weak. The result of sensory evaluation also showed that fame 1 was the best. For the participants with wide faces who were used to traditional eyeglass, the new frame felt "precarious, as though the frame is about to slip," however, the amount of slippage when heads were shaken was confirmed to be about the same as the traditional frame.

  1. M. Kouchi and M. Mochimaru: Analysis of 3-D human face forms for proper sizing and CAD of spectacle frames. Ergonomics, 47:1499-1516, 2004
  2. M. Kouchi, M. Mochimaru: Analysis of 3D Human Face Forms and Spectacle Frames Based on Average Forms, Digital Human Modeling Conference 2002, pp.69-89, 2002
  3. M. Mochimaru, M. Kouchi: Proper Sizing of Spectacle Frames based on 3D Digital Faces, XVth Triennial Congress International Ergonomics Association (IEA2003), 2003