Team
2 students taking the course "Quantitative methods for human-centered design"
Tools
R-Studio, Google Suite
Calipers, scales, reading glasses, jeweler's saw and popsicle sticks for prototyping and trials
Calipers, scales, reading glasses, jeweler's saw and popsicle sticks for prototyping and trials
Summary
This project identified and explored aspects of eye glasses that are affected by users’ body dimensions (anthropometry) and fit preferences in order to develop geometries and sizing schemes that accommodates a wide range of individuals within a target population.
Eyeglass fit is best characterized by the ear-to-temple distance, also called preaurale to sellion distance, interpupillary distance, temple breadth, noise bridge width, and head-breadth at the ears. These landmarks are associated with the arm length of the glasses, the location of the lense centers, the width of the frames, the width of the noise bridge, and the width of the glasses arms respectively.
Considerations for comfortable placement of glasses on one's face
Method
A) Analysis of available data
Linking Design and Anthropometric Dimension
In order to determine how to approach the analysis, the identification of design-critical head and facial features was completed.
+ Frame feature decomposition
+ Landmark selection
+ Landmark selection
Mapping Data to Target Population (Data Imputation)
The second step was evaluation of the target population through use of available anthropometric data sets. Head and facial geometry data needed for the design evaluation was not available for the target population, but was available for a different demographic. To get the needed information for the desired target population, imputation between the two datasets was carried out using correlated landmarks to synthesize a new dataset. Additionally, the synthesized dataset was adjusted to provide the desired male/female ratio.
Landmark selection for the temple-tip width at the ears (distance between the left & right preaurale landmarks) and the lens center reference points (interpupillary distance).
Univariate quantile ranges associated with the desired multivariate accommodation target (middle 95% of the user population).
Multivariate Virtual Fit Testing
Using R-Studio, the ranges of the design-critical facial and head geometries were considered jointly within the chosen population percentile range, in what is know as multivariate virtual fit testing. This step is essential in order to ensure that the final design actually fits the desired percentage of the population (i.e. middle 95% of the population). After the anthropometric ranges were determine “ease” was added to some glasses features, such as a clearance between the temple pieces and the wears’ temples, as an initial step in providing a comfortable fit on the wears’ heads.
Sizing
Given that the ranges of facial and head features varied widely in the target population, a one-size-fits-all approach was determined to be insufficient for providing a high-quality of fit. In order to determine the number of sizes and sizing cutoffs, hypotheses were made about the nature of the user disaccommodation curves with regards to glasses component sizing a fit ranges. In the case of this project, it was determined that three steps per sizing variable would be sufficient to cover the target range of the population. Due to poor correlation between head and facial landmarks from person to person, some of the glasses components would be inappropriate to covary with others from one size to another (i.e. a large head circumference does not correlate to a large interpupillary distance), driving the need for multidimensional sizing, such as a wide or narrow nose version of a small, medium, or large frame.
Preliminary sizing scheme for frame width and temple-tip length.
Preliminary sizing chart based on the scheme
B) User preference evaluation
After determining the initial sizing scheme using the available anthropometric data, some variables needed to be tested based on user fit testing due to the expected weak linkage between some anthropometric landmarks and actual subjects’ preferences and ambiguity on user disaccommodation curves. Specifically, glasses' nose-bridge-width and temple-tip length were selected for the user evaluations. To evaluate both of these variables, a set of prototype glasses were constructed and tested on class-mates. The ultimate goal was to determine general size range preferences and if there were anthropometric cues that could allow for the extrapolation of user fit preferences beyond the data collected in the small class trial.
Results from functional and standard anthropometry are very close, hence proving the sizing derived works
Temple-tip length evaluation (Functional Anthropometric Analysis)
We performed functional anthropometric analysis using regression to predict preferred arm-length of eyeglasses, that is, where to locate the temple tips on the arm of the eyeglasses. Intuitively, we hypothesized that people with larger head length, ear to sellion distance or head circumference would prefer eyeglasses with longer arms. But, there is hardly any correlation with either of these factors. This could be due to -
+ Satisficing
+ Crude measurements
+ Showroom vs Long term comfort
+ Satisficing
+ Crude measurements
+ Showroom vs Long term comfort
We then modelled fit preference using a normal distribution with our dataset’s mean and standard deviation based on the fact that our raw data was approximately normally distributed. We also noticed that there is hardly any difference in the distributions for men and women which helps us conclude that unisex eyeglasses would indeed work for the target population. Further, calculating the cutoffs by optimizing for 90% accommodation, we get a recommended arm length design range of 114mm - 143mm which is very close the sizing range derived from standard anthropometric analysis.
Nose-bridge-width evaluation (Subjective Rating Assessment)
+ Used three frames with three different nose-bridge-widths
+ Measured the nose-breadth of participants and their preference response on a Likert scale
+ Sizing optimization using fit preference logistic regression
+ No correlation between participants “nose-breadth” landmark and preferred glasses nose-bridge-width
In the end, we were not able to identify any correlation between glasses' nose-bridge-width and a subject’s “nose-breadth” landmark. This was likely due to the fact that subjects had different preferences for where they liked to place the nose pads on their nose and the fact that a single landmark is insufficient for describing the complex geometry of the nose.
Nose-bridge-width sizing optimization, using logistic regression. The orange curve represents the probability that the wearer will judge the nose-bridge-width to be too wide and the purple curve represents the probability that the wearer will judge the nose-bridge-width to be too narrow. The solid black curve is the optimization function, representing the probability that the fit will be judged as “just right” by respondents.
Final nose-bridge-width sizing, which exploits the two distinct modes presented for the male and female demographics.
Results
The initial, purely body-dimensions based sizing scheme was appropriate for all glasses dimensions except the nose-bridge-width, which for finer refinement requires further exploration using a greater number of noise landmark points or 3D scanning tools, and a follow-up study of were subjects prefer to place the glasses nose pads.
While head and facial geometry is more complex than other anthropometric dimensions, a reasonable sizing scheme was developed, comprising three principle frame sizes and two width options, for a total of six different frames that need to be produced.
In order to reduce production complexity and cost, while assuring a comfortable fit, it was decided that there was a need to provide a small range of continuous adjustment for the nose-bridge feature, such as flexible wire-supported nose pads.
