An innovative approach to validate AR Glass through software and hardware research
Overview
Rokid has put in tremendous effort and spent millions of dollars to create AR hardware. By combining innovative user-centric research with both software and hardware implementation and development (ergonomics, optics, audio), I helped save hundreds of thousands of dollars that would have been wasted before mass production and increased company exposure and publicity by sharing our research processes and results sequentially AWE events.
Role
Lead UX Researcher, Individual Contributor
Time
January – June 2019
What’s the problem
How do I know if the glass speaker would sound really good?
Goal
Provided recommendations of the mic/speaker models that fit user needs and align with business directions.
Evaluated current structure of mic/speaker ME design and how it should affect the quality of the mic/speaker.
Process
Conducted several rounds of user tests to compare the volume between glass and cellphone speakers. Collaborated with a sound engineer to analyze the sound frequencies produced by both the glass and the cellphone.
Summary
Combined the research results from both quantitative (mouth simulator results) and qualitative (cellphone user tests comparing volume and quality) tests and recommended the most suitable speaker to match users’ needs.
Another UX problem
How do we ensure our AR glass has the best clarity?
Goal
Evaluated subjective clarity of optical solution for AR glass.
Ensured that the user testing results could be applied and translated to optical engineers for revising the optical solutions.
Process
Test #1: Grid Evaluation
Each test participant is given a full-text or full-icon display image and each image is divided by a 4*4 grid. Participants will need to rate the clarity of each area on the grid based off of standardized criteria pre-defined.
Test Participants：
14 people, 7 males and 7 females
Scoring Standard:
Difficult –> Can recognize some color, can’t recognize text
Somewhat Difficult –> Possible to recognize some text, but difficult to correctly identify all words. Causes eye strain.
Easy –> All words are identifiable, but reading them is not easy.
Ideal –> All words are clear and easy to read and does not cause eye strain.
SeenGene
Rokid Glass
Clamping Clip
Interpretation：
For Clamping Clip, most testers think that [the rightmost column] C4 and [lower right corner] R4C3, R4C4 are the most unclear ones.
Clamping Clip R1 performs better than Glass R1, especially with the left C1 column as the best.
Clamping Clip’s other three corners R1C1, R1C4, R4C1 overall (except for the lower right corner R4C4) performs better than Glasses and SeenGene.
SeenGene’s Fov and eye box are bigger and the center area performs better than Clamping Clip and Rokid Glass.
Test #0: Heat map
Each test participant is asked to paint the areas they find unclear with the same color at 10% opacity. The idea is brilliant, but why did it fail? First of all, there was no clear guidance on how they should paint. Sometimes, an area may seem too blurry, prompting them to paint, while at other times, they might feel that they can recognize some text but not all of it. Should they paint or not? Additionally, layering all the opacity can help identify patterns, but how can these be quantified? This leads to our second concern: there is no known way to quantify clear or unclear areas without the aid of a specific tool that we are aware of.
Test Participants：
14 people, 7 males and 7 females
Interpretation：
Clamping Clip has a larger clear area than the glass.
Clamping Clip: the clearest area is 11.5% on average, second to the clearest area is 9.25%, and the hard-hit area is 59%.
Glasses: the clearest area is 8.25% on average, second to the clearest area is 9.5%, and the hard-hit area is 68.75%.
Summary
Setup difference:
The clamping clip is a fixture on the bench that provides lens adjustments before the glasses are taken to assembly. It requires test participants to secure their heads in the alignment part before anything can be seen. The glasses are regular Rokid glasses that have been assembled to reflect the mass-produced pieces.
There are various differences between each piece of testing equipment. For example, the brightness of the glasses is higher than that of the clamping clip. One reason for this is that the clamping clip has exposed lenses, whereas the glasses have lenses inserted inside the frame. To ensure visual similarity, the brightness is assessed with a brightness meter and then adjusted to make sure both viewing experiences are as similar as possible.
Test difference:
Test 0 uses a heat map to depict unclear areas with 10% transparency. Since unclearness cannot be quantified or defined, there may be instances of overrating. Test 1 employs a scoring mechanism, which I proposed as a remedy for Test 0. It divides the same content into a 4×4 grid and helps quantify scores for each area on the grid, minimizing the overrating error that cannot be quantified.
Can we really solve this?
How do we ensure our AR glass would fit various head shapes?
The Problem
To solve weight and misfit issues, we aim to enable people to wear the device for longer periods in greater comfort.
The Weight
How to solve
Understand glasses
Understand heads and faces
Combine
Temple Length
The curvature of the temple arm and length is important. Try to avoid pressing your temple and squeeze your head so hard that will cause you headache.
Glass Curvature
A smooth transition would treat your ear better and will not crush on your ears.
Temple Arm Tip
Make sure it sits on the back of your head bones to provide the support and help balances the front.
Inward Angle
A wide angle would shift the weight to the front of the glass, causing glass to fall on your nose; whereas a narrowed angle should shift the weight to the side of the glass, causing glass to fall on your ears. Neither is comfortable.
Nosepad
The bigger area it touches your nose, the better support it gives to the user.
