An AI-based communication system for motor and speech disabled persons: design methodology and prototype testing

Area of Science:

• Assistive Technology
• Artificial Intelligence
• Computational Linguistics

Background:

• Nonverbal individuals with motor disabilities face challenges in generating written and spoken messages.
• Existing communication aids may lack sophisticated mechanisms for optimizing message generation.
• The integration of knowledge bases and advanced reasoning is crucial for effective assistive communication.

Purpose of the Study:

• To develop an intelligent communication device for the nonverbal, motor disabled.
• To enhance message generation through optimized search strategies.
• To integrate external information and time-varying evidence for improved communication.

Main Methods:

• Development of a knowledge base with grammatical rules and message elements.
• Implementation of a "belief" reasoning scheme integrating external and embedded knowledge.
• Conceptualization of message element search as a path search in a language graph using a frame architecture.
• Augmentation of Bayesian "belief" reasoning (Dempster-Shafer theory) for time-varying evidence.
• Introduction of an "information fusion" strategy for integrating diverse external information.

Main Results:

• A prototype intelligent communication device was developed.
• The system demonstrates an optimized process for message search and generation.
• The belief reasoning scheme effectively utilizes both internal knowledge and external data.
• The information fusion strategy successfully integrates various information sources.
• Experimental testing validated the prototype's functionality.

Conclusions:

• The developed intelligent communication device shows promise in assisting nonverbal individuals.
• The integrated approach of knowledge bases, belief reasoning, and information fusion enhances communication capabilities.
• Further experimental testing and refinement are warranted to optimize the system for widespread use.