iCARE Interaction Assistant: A Wearable Face Recognition System for Individuals with Visual Impairments
Sreekar Krishna, Greg Little, John Black, Sethuraman Panchanathan · 2005 · Proceedings of the 7th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '05) · doi:10.1145/1090785.1090837
Summary
This paper presents the iCARE Interaction Assistant, a wearable assistive device designed to facilitate social interactions for people who are blind or visually impaired by recognizing faces in their vicinity. The system was motivated by input from focus groups who highlighted that people who are blind must often wait for others to initiate social contact and, even after initial contact, lack access to the non-verbal communication cues that sighted people rely on. The prototype consists of a small pinhole analog CCD camera mounted inside the nose bridge of standard eyeglasses, connected to a portable computing element (a tablet PC carried in a backpack) that runs the face detection and recognition algorithms. When the system identifies a known person, it converts the name to speech via the Microsoft Speech Engine and routes the audio to a sound emitter near the user's ear, designed to avoid altering the user's environmental acoustic perception. The system can learn and recognize faces at distances up to 10 feet, with the entire "capture and learn" process taking about 30 seconds per person. The hardware uses a 1/3-inch CCD with 0.2 Lux light sensitivity and a 92-degree field of view, with an Adaptec video digitizer converting the analog signal to compressed AVI over USB.
Key findings
The face recognition pipeline operates in two stages. First, a face detection algorithm based on adaptive boosting identifies regions in video frames containing human faces. Then, two recognition algorithms are applied: Principal Component Analysis (the well-known Eigenfaces method from Turk and Pentland's seminal 1991 work) and a novel method called Distinctive Feature Analysis that recognizes faces based on features that distinguish them from others in the database. To address practical reliability issues where transient environmental variations caused sporadic misrecognition, the system was configured to require five consecutive matching frames before speaking a person's name. The Interaction Assistant is part of the larger iCARE project at Arizona State University's Center for Cognitive Ubiquitous Computing (CUbiC), which aims to develop assistive devices that go beyond navigational aids to support social interaction, learning, and communication. Future work planned includes recognizing and interpreting non-verbal communication cues such as eye contact, facial expressions, emotions, and gestures during conversations.
Relevance
This early wearable face recognition system addresses a social dimension of visual impairment that is often overlooked in assistive technology research — the ability to proactively initiate social contact rather than passively waiting for others to approach. While the 2005 hardware (backpack-mounted tablet PC, wired camera) seems bulky by modern standards, the core concept has since been realized in smartphone apps and smart glasses with far more capable hardware. The research is significant for articulating the social isolation that can result from visual impairment and for framing face recognition not as a surveillance tool but as a social enabler. The focus group methodology that drove the design — listening to what blind users actually need in social situations — is a model for user-centred assistive technology development. The planned future work on interpreting facial expressions and emotions during conversation anticipated capabilities now being explored in modern AI-powered wearables.
Tags: face recognition · wearable technology · blindness and low vision · social interaction · computer vision · assistive technology