At the British Science Festival, visitors used their hands to mix music, change settings and control lighting, without needing a conventional DJ deck.
DJ Air uses camera-based gesture recognition to translate selected movements into controls for a virtual DJ console. The interactive demonstration was developed by postgraduate researchers from the UKRI MINDS CDT and the UKRI AI Centre for Doctoral Training in AI for Sustainability, SustAI.
Turning movement into a digital command
DJ Air tracks a person’s hand and body poses through a camera. Selected gestures are mapped to actions including playing and pausing music, changing volume and crossfading between tracks.
The system uses MediaPipe to detect poses and runs on a laptop. It does not require a specialist motion sensor or physical DJ controller, although it still needs a camera, such as the one built into an ordinary laptop.
That distinction was important at the festival. Visitors were often surprised that the touch-free controls did not depend on specialist camera equipment.
MINDS postgraduate researcher Keniel Peart brought expertise in gesture recognition to the development of the demonstrator. His wider doctoral research explores how AI can interpret detailed hand and body movement using video and radar, with a particular focus on sign-language recognition and human-motion understanding.
DJ Air is not itself a sign-language translation system. Instead, it offers an accessible way to demonstrate some of the computer-vision concepts involved in recognising movement and turning it into an action.
Read more about the gesture-recognition technology behind DJ Air.
From Innovation Camp idea to public demonstration
DJ Air began as an idea developed by MINDS postgraduate researchers Rudra Mutalik and Epifanios Baikas, together with SustAI postgraduate researcher Theodoulos Parpounas, during the 2025 MINDS and SustAI Innovation Camp.
The team’s prototype won the Innovation Camp competition. Further work then developed the idea into a more stable demonstrator that could be used at public events, with Keniel Peart contributing gesture-recognition expertise.
Rudra Mutalik, who undertook a substantial part of the original software development, said the team wanted to explore a more accessible and engaging way to interact with DJ software. The aim was to let the user move and face an audience rather than concentrating on a keyboard or physical controls.
The first prototype used rules to identify gestures from the positions of points detected on the user’s hand. Differences in hand orientation and the limited depth information available from a two-dimensional image created an early technical challenge for the team.
Putting gesture recognition in people’s hands
At the British Science Festival, the music provided an immediate way for visitors to understand the technology. Rather than seeing a model or diagram, they could make a movement and observe the result.
Theodoulos Parpounas, who presented the demonstration at the festival, reported that visitors ranging from young children to older people were eager to try it. He particularly remembered their reaction when they first managed to blend two tracks or control the lighting.
The experience also changed some of the conversations around the display. Some visitors initially questioned where gesture recognition might be useful. After trying DJ Air, they began proposing other situations in which similar technology might be applied.
“Public engagement can help researchers broaden their perspective,” said Theodoulos Parpounas. “By presenting their work to a wide audience and engaging in active discussions, it can help them verify the importance of what they do, as well as identify aspects of their work that they did not consider before.”
Exploring wider uses for gesture-controlled technology
Music provides an engaging demonstration, but the underlying interaction method has possible uses beyond a virtual DJ console.
Camera-based gesture recognition could contribute to accessible interfaces, contact-free controls, robotics, rehabilitation monitoring and virtual or augmented reality. It could also support future research into systems that respond to sign language or other forms of human movement.
These are potential applications rather than outcomes already delivered by DJ Air. Systems designed for wider use would also need to address technical and social questions, including recognition accuracy in complex environments, variation between people’s gestures and the privacy implications of using an active camera.
For Keniel Peart, the project provides a public-facing example of how computers can be taught to recognise movement. It also connects an immediately understandable activity with wider MINDS research into computer vision, gesture recognition and more accessible ways of interacting with technology.
Watch DJ Air in action
See Keniel Peart and Theodoulos Parpounas demonstrate DJ Air and explain how gesture recognition can turn movement into a digital command.
Discover more
Learn more about Keniel Peart’s research into sign-language recognition, human-motion understanding, computer vision and radar sensing: