A collaborative research project has developed a lightweight, wearable system that is also easy to use, according to a new study. — SCMP
Chinese scientists say they have developed a wearable artificial intelligence system that can help visually impaired people to navigate their way around the world.
An AI algorithm analyses real-time footage of the environment and gives the user concise directional prompts via bone conduction headphones, according to a paper published on Monday in the peer-reviewed journal Nature Machine Intelligence.
Artificial skin sensory motors on each wrist monitor the user’s surroundings and vibrate an alert if they detect potential obstacles on either side, it said.
The system was developed by engineers from Shanghai Jiao Tong University, Shanghai Artificial Intelligence Laboratory, East China Normal University, and the Hong Kong University of Science and Technology in collaboration with researchers from the State Key Laboratory of Medical Neurobiology at Fudan University.
“This research paves the way for user-friendly visual assistance systems, offering alternative avenues to enhance the quality of life for people with visual impairment,” they wrote.
Lead author Gu Leilei, an associate professor with Shanghai Jiao Tong University’s School of Electronic Information and Electrical Engineering, said that the AI-powered system was designed to optimise the user’s experience through intuitive cues.
“Lengthy audio descriptions of the environment can overwhelm and tire users, making them reluctant to use such systems,” Gu said.
“Unlike a car navigation system with detailed directions, our work aims to minimise AI system output, communicating information key for navigation in a way that the brain can easily absorb.”
Gu added that the team also used lightweight and compact hardware to create a portable system, which was tested by 20 visually impaired people who were invited to take part in the study.
According to the paper, the feedback from the 20 participants showed “the system’s reliable functionality and high usability”.
Gu said users navigating an indoor setting would take 10 to 20 minutes to familiarise themselves with the system, walking around the venue several times.
After the user sets a target via voice commands, the AI algorithm processes footage from a camera mounted between the person’s eyebrows to estimate the destination’s orientation and determine an obstacle-free route.
The algorithm, which runs locally on the device, has been trained to recognise 21 objects placed at different distances and angles, according to the paper.
These include furniture, such as beds, tables, chairs and doors, as well as kitchen items, from sinks to utensils and foodstuffs, along with home appliances like televisions, and of course other people. The training can be expanded to cover more items, the paper said.
In addition to the system’s audio cues, the ultrathin artificial skin sensors on the wrist can detect the distance between the user’s hands and nearby objects. If the distance to a wall on the user’s left falls below the customised threshold, the corresponding wrist sensor will vibrate, prompting a move to the right, for example.
The ultrathin artificial skin could also help users to reach and grasp objects by guiding them with information about the distance between their hand and the target, according to the study.
Gu said the research team will continue working to enhance the system’s accuracy and safety when navigating outdoor environments, which are more complex than indoor ones. – South China Morning Post
