A Google Glass App to Help the Blind in Small Talk
Mohammad Iftekhar Tanveer, Mohammed Ehsan Hoque · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS) · doi:10.1145/2661334.2661338
Summary
This demo paper presents a Google Glass prototype that helps blind individuals initiate small talk by automatically analysing a scene to detect the number of people present, their approximate ages, and gender distribution, then synthesising this information as speech feedback through the Glass earbuds. The motivation is that initiating conversation — particularly with strangers or groups — requires social cues that sighted people pick up visually: how many people are nearby, their approximate age and gender (which influence appropriate conversation topics and approach), and their spatial arrangement. Blind individuals lack these cues, which can make social situations awkward or intimidating. The system works by capturing an image (via button press or double-tap gesture on the Glass), uploading it to the cloud or a nearby local computer for processing, performing face detection and facial feature analysis using machine learning, and returning a synthesised audio summary. For example, given a group of four people, the system might say: "Group of 4. Age range 15 to 22. Average age 18. 50% male." The authors chose Google Glass over dedicated assistive devices to reduce social stigma — prior research by Kane et al. showed that using mainstream consumer devices rather than visibly assistive technology attracts less attention, reduces cost, improves mobility, and increases social acceptance. The paper also contrasts the automated computer vision approach with crowdsourcing-based systems like VizWiz, noting that crowd workers may be inconsistent, non-real-time, expensive at scale, and raise privacy concerns when sharing images with strangers.
Key findings
The prototype demonstrates the technical feasibility of real-time social scene analysis on a wearable platform, providing three key pieces of information: group size, age distribution, and gender ratio. The cloud-based processing architecture offloads computationally intensive face detection and feature analysis from the Glass hardware, which has limited processing power and battery life. The system addresses a specific social accessibility need — the ability to assess a social scene before engaging — that is distinct from other wearable assistive applications like navigation or object recognition. The authors acknowledge that the current implementation is limited to basic demographic information and does not yet include facial expression recognition, identity recognition, or analysis of body language and spatial arrangement. The paper is explicit that real-world validation with blind users in actual social scenarios remains future work, positioning this as a technical proof of concept rather than a validated assistive tool.
Relevance
This paper, alongside the Expression system presented at the same conference, represents an early exploration of wearable social assistive technology for blind users. While Expression focused on reading facial expressions during ongoing conversations, this system addresses the preceding challenge: assessing a social scene before initiating interaction. Together, they illustrate the range of visual social information that blind people miss and that technology could potentially provide. For accessibility practitioners, the paper raises important ethical considerations that have become more prominent since 2014: automated age and gender classification involves making assumptions about people's identity based on appearance, which intersects with concerns about AI bias, facial recognition accuracy across demographics, and the ethics of classifying strangers without consent. The choice to use consumer wearables (Google Glass) rather than visibly assistive devices reflects the social model of disability — reducing stigma by using mainstream technology. Though Google Glass was discontinued as a consumer product, the underlying concept of scene-level social awareness through smart glasses remains actively pursued as AR glasses technology advances.
Tags: blindness · Google Glass · wearable technology · computer vision · face detection · social interaction · social accessibility