With unreliable internet and power, rural terrains and unique speaking patterns, Africa would need its own unique AI solutions
Artificial intelligence (AI) is no longer limited to chatbots, automated writing tools or online assistants. It is being built into wheelchairs, computer interfaces and other physical devices. These systems can understand their surroundings, respond to the user’s movements and help people with severe physical disabilities regain greater control. For people with high-level spinal cord injuries, this could be life changing.
Cost obstacles
AI-powered mobility devices and smart control systems can cost thousands of US dollars. This can translate into tens or even thousands of rands for South Africans. Many assistive devices developed overseas are designed for environments with smooth pavements, reliable electricity, affordable internet and stable mobile networks.
These conditions do not reflect the daily experiences of many South Africans. Our devices may need to operate during power cuts, function without constant internet access and move across uneven roads or unpaved terrain. They need to be affordable to repair and simple enough to maintain locally.
Lightweight solutions
Brain-computer interfaces often receive the most international attention. These systems allow users to control devices through brain activity. Although promising, it remains extremely expensive and medically complex. Less invasive systems may offer a more practical path.
Lightweight AI uses a camera to follow eye movements, detect facial gestures or recognise intentional blinks. It can run on small computers and affordable processors without requiring complex surgery or specialist medical equipment.
Researchers at Nelson Mandela University, for example, developed an intelligent wheelchair control system using a five- megapixel infrared camera and a Raspberry Pi 4 computer. The system uses facial-tracking software to identify hundreds of points on the user’s face. It follows the direction of the eyes and measures intentional blinks to recognise commands. The system reportedly achieved a command recognition accuracy of 98.71 percent, even in very dark conditions.
It was built using relatively affordable, widely available components, which shows that advanced assistive technology does not always require expensive custom hardware.
On-device solutions
A smart wheelchair can use cameras and sensors to identify obstacles, dangerous edges, uneven ground or possible drop- offs. It can then make small adjustments to the user’s steering input to reduce the risk of a collision or fall while the user still decides where to go. AI acts as a safety layer.
This type of support is especially valuable in environments where paths may be damaged, pavements may be missing and wheelchair users must constantly navigate unexpected hazards. Computer vision systems can run on small processors such as the NVIDIA Jetson range.
The most valuable feature of these systems may be their ability to continue working without internet. A device that depends entirely on online servers become unreliable when data runs out, the network fails or power cuts out. It can mean losing access to an essential part of daily life.
This is why offline and on-device AI is so important for Africa. With on-device processing, the camera, microphone or sensors send information directly to a small computer inside the device. The AI makes decisions locally instead of sending every command to a distant server. This allows the system to respond faster, protect the user’s privacy and continue working without a constant internet connection.
Understanding local speech
Local voice technology is another important area of development. Many international voice assistants struggle to understand African accents, languages and speech patterns. This can make voice control unreliable. African technology companies are working to change this.
Lelapa AI’s Vulavula platform is being developed to support African languages and speech. Intron Health has worked on speech- recognition systems designed to understand African accents more accurately. These tools could eventually support hands-free control of computers, wheelchairs, phones and smart home devices.
Beyond the prototype
South Africa already has the talent and technical ability to develop these systems. Researchers at Wits University have shown how affordable consumer electronics, including sensors from Nintendo Wii controllers, can be adapted for hands- free computer control. Projects such as the Dassie X rural power wheelchair show that mobility devices can be designed around rough terrain, limited transport and the need to fit into public minibus taxis.
The next challenge is moving from research prototypes to reliable products. Organisations such as Uku’Hamba show how local manufacturing can combine social impact with practical innovation. By using 3D printing and recycled materials, local enterprises can produce adaptive products while creating technical training and employment opportunities for young people with disabilities.
Universities should work more closely with manufacturers, disability organisations and social enterprises. Telecommunications and technology companies also have a role to play. They could subsidise locally produced assistive hardware, support the development of accessibility software and reduce data costs for essential digital services.
Keeping the user central
Assistive technology should always be developed with people with disabilities rather than for them. Lived experience can reveal problems and guide better decisions about comfort, safety, maintenance and everyday use. Organisations like the QuadPara Association of South Africa can help researchers and manufacturers with testing, training, distribution and long-term support.
Accessibility tools should not be treated as optional luxury products. For many people, they are a necessity. South Africa does not need to wait for foreign technology companies to reduce their prices. We can combine affordable AI with locally designed hardware that works offline, survives unstable infrastructure and reflects the realities of our communities. We already have the researchers, engineers, disability advocates and local innovators required to begin.
