Team:

Natalia Elyssa Chan is an undergraduate student from the National University of Singapore with interests in AI, accessibility technology, and human centred design. She initiated and led the development of LIA, an assistive AI project aimed at helping visually impaired users identify household objects and products through voice interaction and computer vision.

​

Her role in the project included defining the product direction, designing the system logic, prototyping the accessibility workflow, and developing the final project demonstration and presentation materials. During development, the project was also further refined through collaborative prototyping opportunities and external demonstration settings.

​

Project Overview:

LIA (Listen, Identify, Assist) is an assistive AI prototype designed to help visually impaired users interact more independently with everyday household objects and products.

​

Using a phone camera and voice interaction, users can point at an object, ask a spoken question such as “What is this?” or “Can I take this medicine?”, and receive a spoken response from the system. The project combines computer vision, speech recognition, speech synthesis, and structured product lookup to create a simple and accessible user experience.

​

A key focus of the project was transparency and safety. Rather than presenting AI answers with false certainty, the system was designed to recognise uncertainty and recommend human assistance in situations where confidence is low or where questions involve potentially sensitive products such as medication or chemicals.

 

Why the team applied for CCSGP Fellowship:

The project was strongly aligned with CCSGP’s mission of using technology for meaningful social impact. LIA was created to explore how accessible AI systems could improve independence and confidence for visually impaired users in everyday situations.

The CCSGP project provided an opportunity to experiment with practical assistive technology ideas, explore responsible AI design, and develop a prototype focused on accessibility, trust, and real world usefulness.

 

About the Project:

LIA was developed as a prototype assistive AI system focused on helping visually impaired users perform everyday household tasks more independently.

​

The system allows users to interact entirely through voice and camera input. A user can point their phone camera at an object, ask a spoken question, and receive a spoken response generated through AI powered image understanding and product lookup systems.

​

The prototype combines several technologies into a single workflow, including:

• Computer vision for object recognition

• Barcode based product identification

• Speech to text voice interaction

• Text to speech spoken responses

• Confidence based transparency logic

• Human assistance escalation recommendations

• ​

One of the project’s main goals was addressing the problem of false confidence in AI systems. Instead of always presenting answers as certain, the system was designed to surface uncertainty clearly and encourage human assistance in situations where incorrect information could pose safety risks.

 

The project later received external recognition during a collaborative hackathon demonstration, where the prototype was commended for its accessibility focused design and thoughtful approach toward trustworthy assistive technology.

​

What the team learned from the CCSGP experience:

The project provided valuable insight into the challenges of designing technology for accessibility and social impact. One of the most important lessons learned was that assistive AI systems must prioritise trust and transparency in addition to technical performance.

Through the development process, it became clear that confidence and uncertainty handling are especially important when users rely entirely on spoken responses. This shaped many of the project’s design decisions, including the implementation of confidence based guidance and volunteer escalation recommendations.

​

The project also provided experience in rapid prototyping, accessibility focused product design, AI integration, and communicating technical ideas through demonstrations and public presentations.

​

Screenshots:

​