Lessons Learned from Developing a Privacy-Preserving Multimodal Wearable for Local Voice-and-Vision Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tussa, Yonatan, Heredia, Andy, Roy, Nirupam
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909922235514880
author Tussa, Yonatan
Heredia, Andy
Roy, Nirupam
author_facet Tussa, Yonatan
Heredia, Andy
Roy, Nirupam
contents Many promising applications of multimodal wearables require continuous sensing and heavy computation, yet users reject such devices due to privacy concerns. This paper shares our experiences building an ear-mounted voice-and-vision wearable that performs local AI inference using a paired smartphone as a trusted personal edge. We describe the hardware-software co-design of this privacy-preserving system, including challenges in integrating a camera, microphone, and speaker within a 30-gram form factor, enabling wake word-triggered capture, and running quantized vision-language and large-language models entirely offline. Through iterative prototyping, we identify key design hurdles in power budgeting, connectivity, latency, and social acceptability. Our initial evaluation shows that fully local multimodal inference is feasible on commodity mobile hardware with interactive latency. We conclude with design lessons for researchers developing embedded AI systems that balance privacy, responsiveness, and usability in everyday settings.
format Preprint
id arxiv_https___arxiv_org_abs_2511_11811
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lessons Learned from Developing a Privacy-Preserving Multimodal Wearable for Local Voice-and-Vision Inference
Tussa, Yonatan
Heredia, Andy
Roy, Nirupam
Human-Computer Interaction
Systems and Control
Audio and Speech Processing
Image and Video Processing
Many promising applications of multimodal wearables require continuous sensing and heavy computation, yet users reject such devices due to privacy concerns. This paper shares our experiences building an ear-mounted voice-and-vision wearable that performs local AI inference using a paired smartphone as a trusted personal edge. We describe the hardware-software co-design of this privacy-preserving system, including challenges in integrating a camera, microphone, and speaker within a 30-gram form factor, enabling wake word-triggered capture, and running quantized vision-language and large-language models entirely offline. Through iterative prototyping, we identify key design hurdles in power budgeting, connectivity, latency, and social acceptability. Our initial evaluation shows that fully local multimodal inference is feasible on commodity mobile hardware with interactive latency. We conclude with design lessons for researchers developing embedded AI systems that balance privacy, responsiveness, and usability in everyday settings.
title Lessons Learned from Developing a Privacy-Preserving Multimodal Wearable for Local Voice-and-Vision Inference
topic Human-Computer Interaction
Systems and Control
Audio and Speech Processing
Image and Video Processing
url https://arxiv.org/abs/2511.11811