Fieldlab UPPS

Investigating the Value of Statistical Ear Shape Modelling in the Earbud Personalisation Process

31.1.2024
Dopple Product Factory B.V.

Summary

In the UPPS-project, Dopple, together with Delft University (IDE), has worked out the question of how to make fully personalised earbuds. We started by building a database. This database consists of a 3D scan of the ear canal and concha, a full 3D head scan, and a photo of both ears. A statistical 3D ear model was generated from the 3D ear scan data. With a master thesis, a link was made between different ear scan techniques and the modelling of the ear fit of the client by using Grasshopper software. Placing of the electronics in the ear scan data is done manually. With the course Computational Design For Digital Fabrication, various methods have been tested to modulate the ear fit and positioning of the (Dopple) electronics, with the intention to prepare for an automated process

Problem Definition

The project aims to create a 3D statistical ear shape model for a representative population, integrating potentially augmented data like ear photos, to address the variability in ear shapes during customer measurement and design automation for custom-fit earbuds. This model will be assessed for its effectiveness in estimating customer ear geometry efficiently, extracting necessary ear dimensions automatically, and translating these geometries into precise earbud design parameters, thereby streamlining the design process. Through this investigation, the project seeks to demonstrate the significant benefits of employing a statistical ear shape model in enhancing the personalization and production efficiency of consumer electronics, particularly in the development of tailored wireless earbuds by Dopple.

UPPS Workflow Description

Collect Phase

Collecting

Dopple and Fieldlab UPPS combine their expertise to develop a statistical ear shape model. They focus on gathering 3D ear shapes and customer photos. Dopple collected customer ear shapes and photos (546 in total), far exceeding their target of 220. This was achieved through scanning sessions at both TU Delft and Dopple. The collected data were then processed into STL files of ear geometry and PNG photos.

Equipment

The project utilised innovative equipment, such as smartphones for capturing ear shapes, alongside existing 3D scanning technologies. These tools aided in efficiently gathering the necessary data without the need for customers to visit an audiologist.

Analyse Phase

Selection

The selection of collected data was carefully considered, aiming to develop a representative statistical model of ear shape. This model was enriched with additional data, such as ear photos, which was crucial for the accuracy of the model.

Comparison

The collected data, including 3D scans and ear photos, were compared with each other to construct an accurate database. This process was essential for developing a reliable statistical model of ear shape.

Design Phase

Parametric Modelling

The development of the statistical ear model facilitated parametric modelling. This model was adapted based on the collected data, making the design process more efficient and avoiding the need for designers to start from scratch repeatedly.

Co-creation

The project encouraged co-creation by integrating customer preferences into the design of personalised earbuds. This process was supported by the collaboration between Dopple and Fieldlab UPPS, allowing users and developers to collaboratively work towards the ideal product.

Produce Phase

Use Phase

Conclusion

1. The statistical ear shape model is successfully generated at TU-Delft based on the ear database built by Dopple and placed in the DINED anthropometric database. The actual status is that this statistical ear shape model is used to improve the concha data obtained by a smartphone camera (master thesis work). The final result is good enough for personal fit product design in the concha. The ear canal is not sufficiently detailed yet for final product design. The pinna has not been considered yet.

2. For the concha data, we can create good personal fit with the help of smart phone camera, customer ear scan and statistical ear shape model derived from the new ear database. It is not completely automated yet, but we made great progress.

3. With the help of TU-Delft (course for Computational Design For Digital Fabrication (CfD)) the study shows good possibilities to automate the best location for the technical components in the custom ear product. Further steps have to be made to proof that this operation of automated placement for the core components will be successful. 

No items found.