Page 743 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

those in remote or underserved areas—to learn lip-reading independently via online platforms.
This helps address the critical shortage and uneven distribution of educational resources,
contributing to inclusive and equitable access to quality education and lifelong learning
opportunities. 4.9: Accessibility

By replacing conventional, human-led teaching methods with AI-driven tools, the solution
significantly reduces educational costs while offering standardized, repeatable learning
experiences. It also supports the advancement of education infrastructure and digital learning
capabilities, particularly for learners in areas where special education resources have long
been scarce.

Reduced Inequalities: The project tackles communication-related exclusion by equipping deaf
individuals with AI-based tools to enhance their speech and lip-reading abilities. This supports
more active participation in social life and helps reduce disparities caused by physiological
challenges. With 343 individuals already enabled to speak using the system, it has proven
effective in increasing confidence and enabling integration into broader society.

By offering free access to deaf schools and rehabilitation centers—especially in resource-
limited regions—the initiative ensures that vulnerable populations receive targeted support.
For example, deaf individuals in rural areas of China can independently engage in learning
without depending on in-person instruction, narrowing the gap in access between urban and
rural areas.

In contrast to traditional systems that focus solely on helping hearing people understand the
deaf (e.g., sign language interpreters), this solution empowers deaf individuals to express
themselves vocally, fostering two-way communication. This shift toward vocal empowerment
plays a vital role in building a more inclusive and communication-accessible society.

2�3 Future Work

Data Collection: n:

We will expand Chinese lip-reading datasets to enhance AI model accuracy in Chinese visual
speech recognition. This will improve environmental adaptability (e.g., strong backlighting/
low-light conditions), reduce physical space requirements for lip-reading training, and ensure
equal access to educational resources for deaf communities in underdeveloped regions.

Model Development: t:

Through lightweight model optimization for local deployment on mobile devices, we aim to:

- Reduce on-device feedback latency

- Minimize reliance on stable internet connections

This technological advancement will optimize usability in remote areas and low-signal scenarios,
guaranteeing equitable access to efficient learning support.

Standardized Educational Framework: k:

We will refine system functionalities and pedagogical methods through practical implementation,
ultimately establishing a standardized deaf education framework based on lip-reading and

707