Human-inspired sound environment recognition system for assistive vehicles.

Area of Science:

• Robotics and Artificial Intelligence
• Human-Computer Interaction
• Assistive Technology

Background:

• The human auditory system processes environmental information faster than visual or tactile systems, enabling quicker responses.
• Assistive robotic devices like wheelchairs require robust environment recognition for safe navigation.
• Existing sensors (LiDAR, vision) have limitations due to environmental constraints like lighting and object color.

Purpose of the Study:

• To implement a sound-based approach for environment recognition in assistive robotic devices.
• To enhance the safety of robotic wheelchairs and cars by utilizing auditory cues.
• To classify environments based on sound to protect users, especially those with disabilities, in community-based physical activities.

Main Methods:

• A neural network was developed to classify up to 15 different environments based on sound stimuli.
• The system was designed considering the International Classification of Functioning (ICF) framework for environment factors relevant to people with disabilities.
• Real-time outdoor experiments were conducted with seven volunteers using an assistive vehicle.

Main Results:

• The sound-based environment classification achieved accuracy rates between 84% and 93%.
• Classified environmental data was used to constrain assistive vehicle navigation, enhancing user protection.
• Statistical validation and comparison with prior work confirmed the system's efficacy.

Conclusions:

• The sound-based system effectively provides environmental descriptions, particularly for vulnerable situations identified by the ICF.
• Volunteers reported comfort and confidence in the system's performance and its ability to adjust vehicle speeds in risky environments.